Different associations between amyloid-βeta 42, amyloid-βeta 40, and amyloid-βeta 42/40 with soluble phosphorylated-tau and disease burden in Alzheimer’s disease: a cerebrospinal fluid and fluorodeoxyglucose-positron emission tomography study

Background Despite the high sensitivity of cerebrospinal fluid (CSF) amyloid beta (Aβ)42 to detect amyloid pathology, the Aβ42/Aβ40 ratio (amyR) better estimates amyloid load, with higher specificity for Alzheimer’s disease (AD). However, whether Aβ42 and amyR have different meanings and whether Aβ40 represents more than an Aβ42-corrective factor remain to be clarified. Our study aimed to compare the ability of Aβ42 and amyR to detect AD pathology in terms of p-tau/Aβ42 ratio and brain glucose metabolic patterns using fluorodeoxyglucose-positron emission tomography (FDG-PET). Methods CSF biomarkers were analyzed with EUROIMMUN ELISA. We included 163 patients showing pathological CSF Aβ42 and normal p-tau (A + T −  = 98) or pathological p-tau levels (A + T +  = 65) and 36 control subjects (A − T −). A + T − patients were further stratified into those with normal (CSFAβ42 + /amyR −  = 46) and pathological amyR (CSFAβ42 + /amyR +  = 52). We used two distinct cut-offs to determine pathological values of p-tau/Aβ42: (1) ≥ 0.086 and (2) ≥ 0.122. FDG-PET patterns were evaluated in a subsample of patients (n = 46) and compared to 24 controls. Results CSF Aβ40 levels were the lowest in A − T − and in CSFAβ42 + /amyR − , higher in CSFAβ42 + /amyR + and highest in A + T + (F = 50.75; p < 0.001), resembling CSF levels of p-tau (F = 192; p < 0.001). We found a positive association between Aβ40 and p-tau in A − T − (β = 0.58; p < 0.001), CSFAβ42 + /amyR − (β = 0.47; p < 0.001), and CSFAβ42 + /amyR + patients (β = 0.48; p < 0.001) but not in A + T + . Investigating biomarker changes as a function of amyR, we observed a weak variation in CSF p-tau (+ 2 z-scores) and Aβ40 (+ 0.8 z-scores) in the normal amyR range, becoming steeper over the pathological threshold of amyR (p-tau: + 5 z-scores, Aβ40: + 4.5 z-score). CSFAβ42 + /amyR + patients showed a significantly higher probability of having pathological p-tau/Aβ42 than CSFAβ42 + /amyR − (cut-off ≥ 0.086: OR 23.3; cut-off ≥ 0.122: OR 8.8), which however still showed pathological values of p-tau/Aβ42 in some cases (cut-off ≥ 0.086: 35.7%; cut-off ≥ 0.122: 17.3%) unlike A − T − . Accordingly, we found reduced FDG metabolism in the temporoparietal regions of CSFAβ42 + /amyR − compared to controls, and further reduction in frontal areas in CSFAβ42 + /amyR + , like in A + T + . Conclusions Pathological p-tau/Aβ42 and FDG hypometabolism typical of AD can be found in patients with decreased CSF Aβ42 levels alone. AmyR positivity, associated with higher Aβ40 levels, is accompanied by higher CSF p-tau and widespread FDG hypometabolism. Supplementary Information The online version contains supplementary material available at 10.1186/s13195-023-01291-w.


Background
According to the amyloid cascade hypothesis, Alzheimer's disease (AD) develops due to amyloid peptides (Aβ) deposition in senile plaques, followed by the accumulation of hyperphosphorylated tau proteins (p-tau) in tangles, ultimately leading to neuronal degeneration and cognitive decline.In 2018, the National Institute on Aging-Alzheimer's Association (NIA-AA) reported that the highest probability of identifying AD pathology in vivo is achieved by combining markers of Aβ (A) and p-tau (T) pathology [1].Indeed, it has been demonstrated that the co-presence of A + and T + status, as well as combining them in the CSF p-tau/Aβ 42 ratio, provides excellent predictive power for the presence of AD pathology at autopsy [2,3].However, when there is no concordance between A and T, and notably when the CSF levels of Aβ 42 are abnormal (A +) without concomitant abnormal values of p-tau (T −), we encounter an ambiguous biological profile, labeled "AD pathological changes, " which is still considered part of the Alzheimer's continuum.Given the crucial role of the "A" status in this biological condition, inconsistencies among different amyloid biomarkers, from cerebrospinal fluid (CSF) or imaging, can represent a significant limitation in supporting a possible diagnosis of AD, thus fueling the need to improve their accuracy.
In this regard, it has been demonstrated that the addition of CSF Aβ 40 to CSF Aβ 42 levels in the Aβ 42 /Aβ 40 ratio (amyR) could account for inter-individual variations in Aβ production, thereby enhancing the diagnostic performance of this biomarker [4][5][6][7][8].Furthermore, amyR seems to better reflect the total amount of amyloid brain deposition [9][10][11][12].Nevertheless, Vromen and colleagues also found that the decrease of CSF Aβ 42 alone shows excellent accuracy in detecting autopsy-confirmed AD [13].Together, these data support the idea that amyR and Aβ 42 may provide different kinds of information on the underlying AD pathophysiology.
In the present study, we aimed to determine whether CSF Aβ 40 has a specific meaning, which enables amyR to better estimate the AD-related burden, and whether pathological CSF Aβ 42 in the presence of normal amyR can still identify AD pathology.To address this issue, we focused on the A + T − biological condition, stratifying patients with decreased CSF Aβ 42 into those with normal (CSFAβ 42 + /amyR −) or pathological amyR (CSFAβ 42 + / amyR +).We assessed differences and overlaps with healthy control subjects (A − T −) and full-blown AD patients (A + T +) in terms of disease burden measured as CSF p-tau/Aβ 42 [2,14] and cerebral glucose hypometabolism, evaluated with fluorodeoxyglucose-positron emission tomography (FDG-PET), whose specific topographical patterns have achieved an increasingly supportive role in the diagnostic algorithm of AD [15][16][17].

Subjects' enrolment
Between 2017 and 2021, we evaluated 250 patients at the Memory Clinic of the "Policlinico Tor Vergata" in Rome.The criteria for retrospective inclusion were as follows: (1) a complete diagnostic workup, including standardized neurological examination, laboratory testing, MRI imaging, FDG-PET scan, neuropsychological assessment, APOE genotyping, and CSF analysis, and (2) fulfillment of the diagnostic criteria for dementia [18] or mild cognitive impairment due to AD [19].Exclusion criteria were as follows: (1) presence of other neurological or psychiatric diseases or medical conditions potentially associated with cognitive deficits; (2) major comorbidities such as oncological history, systemic inflammatory conditions, and organ failure; (3) prominent cortical or subcortical infarcts; (4) history of drug or alcohol abuse; (5) use of antipsychotics, antidepressants, or serotonergic drugs.
Eventually, we selected 163 patients belonging to the Alzheimer's continuum (ADc) according to their biomarker profile [1].Further stratification into AT groups was performed according to the presence of decreased CSF levels of Aβ 42 (A) and increased CSF p-tau (T) (see Additional file 1).
Furthermore, we identified 36 controls among inpatients from the Neurology Unit of Policlinico Tor Vergata who had undergone a complete neurological evaluation, brain CT, and lumbar puncture for diagnostic purposes and for which the presence of any primary neurological disease had been excluded.All CSF analyses showed normal cell counts and biomarker profiles.

CSF sampling/analysis and APOE genotyping
All lumbar punctures were performed between 8 and 10 am.An 8 ml CSF sample was collected for each patient in polypropylene tubes, 2 ml were used for routine biochemical analysis, 6 ml were centrifuged at 2000 g at + 4 °C for 10 min, and frozen at -80 °C.All samples were processed according to the manufacturer's instructions and to laboratory standard operating procedures.
Moreover, 24 subjects (male, 10; female, 14; mean age, 66.56 ± 10.44 years) undergoing FDG-PET/CT for other reasons were enrolled as part of a control group (CG) upon showing no signs of hypometabolism suggestive of neurodegenerative disorders nor other cerebral abnormalities, as assessed by visual reads (A.C).All subjects had an MRI performed within 14 ± 4 days before PET/CT examination, showing absence of brain alterations.An experienced neurologist (A.M.) evaluated all participants to assess the absence of clinical signs of cognitive decline.Patient and control selection strategies for the sub-study are summarized in Additional file 2.

FDG-PET/CT scanning and acquisition
The same scanning and acquisition protocol was used for both patients and CG.All subjects were injected with intravenous FDG (dose range 185-295 megabecquerels) and hydrated with 500 ml of saline (0.9% sodium chloride).PET/CT acquisition started 30 ± 5 min after FDG injection and lasted for 10 min in all subjects.The reconstruction parameters were as follows: ordered subset expectation maximization, four subsets and 12 iterations; matrix 256 × 256; full width at half maximum (FWHM): 5 mm.

Data management and statistical analysis CSF biomarkers analysis
All continuous variables were expressed as mean ± standard deviation.The Kruskal-Wallis test and Dunn's post hoc analysis were used for multiple comparisons.Categorical variables were analyzed using Pearson's chisquare test.We computed a linear regression model with age and sex as covariates to study the association between CSF amyloid biomarkers and CSF p-tau levels.We then performed a robust locally weighted regression analysis [21] and plotted CSF biomarkers levels as a function of amyR, using CSF biomarkers values converted to z-scores by subtracting the mean and dividing by the standard deviation of the control group (A − T −).Eventually, logistic multivariate regressions were performed to evaluate the odds of pathological p-tau/Aβ 42 values according to amyR status, accounting for clinical and demographic factors (age, sex, APOE, diabetes mellitus, hypertension, dyslipidemia).
All statistical analyses were performed using Stata-Corp© (Stata Statistical Software: Release 13.College Station, TX: StataCorp) and GraphPad Prism version 9.3.1 (GraphPad Software, San Diego, California, USA).All results were computed with two-tailed significance; p < 0.05 were considered significant.

FDG-PET analysis
First, all FDG-PET scans were visually evaluated in standardized transaxial, coronal, and sagittal plans by an experienced nuclear medicine physician (A.C) and interpreted according to the latest EANM guidelines for FDG-PET imaging [22].We used Statistical Parametric Mapping 12 (SPM12) in MATLAB 2018a (https:// www.fil.ion.ucl.ac.uk/ spm/ softw are/ spm12/) to perform statistical analysis.FDG-PET data were converted from DICOM to Nifti format using Mricron software (https:// www.nitrc.org/ proje cts/ mricr on) and then subjected to a normalization process.A bias regularization was applied (0.0001) to limit biases due to smoothness, spatially varying artifacts that modulate the intensity of the image and that can impede automating processing of images.FWHM of Gaussian smoothness of bias (to prevent the algorithm from trying to model out intensity variation due to different tissue types) was set at 60 mm cut-off; tissue probability map implemented in SPM12 was used (TPM.nii).A mutual information affine registration with the tissue probability maps [23] was used to achieve approximate alignment to ICBM space template-European brains [24,25].Warping regularization was set with the following 1 by 5 array (0, 0.001, 0.5, 0.05, 0.2); smoothness (to cope with functional anatomical variability that is not compensated by spatial normalization and to improve the signal-tonoise ratio) was set at 5 mm; sampling distance (that encodes the approximate distance between sampled points when estimating the model parameters) was set at 3. We applied an 8-mm isotropic Gaussian filter to blur the individual variations (especially gyral variations) and to increase the signal-to-noise ratio.We used the following parameters and post-processing tools before regression analysis was applied: global normalization (that scales images to a global value) = 50 (using proportional scaling); masking threshold (that helps to identify voxels with an acceptable signal in them) was set to 0.8; transformation tool of statistical parametric maps into normal distribution; correction of SPM coordinates to match the Talairach coordinates, subroutine implemented by Matthew Brett (http:// www.mrc-cbu.cam.ac.uk/ Imagi ng).Brodmann areas (BA) were identified at a range from 0 to 3 mm from the corrected Talairach coordinates of the SPM output isocenter, using a Talairach client available at http:// www.talai rach.org/ index.html.As proposed by Bennett et al. [26], SPM t-maps were corrected for multiple comparisons using the false discovery rate (p ≤ 0.05) and corrected for multiple comparisons at the cluster level (p ≤ 0.001).The level of significance was set at 100 contiguous voxels (5 × 5 × 5 voxels, i.e., 11 × 11 × 11 mm).The following voxel-based comparisons were assessed: CG vs. CSFAβ 42 + /amyR − , CSFAβ 42 + /amyR − vs. CSFAβ 42 + / amyR + , and CSFAβ 42 + /amyR + vs.A + T + .We used a full factorial design implemented in SPM12 to test the hypothesis that differences among groups exist overall.Comparisons between groups were performed using the two-sample t-test model of SPM12.For both analyses, age and sex were used as covariates, and the threshold was set at p < 0.001 (p < 0.05 FWE corrected at the cluster level).

Association between amyloid biomarkers and p-tau CSF levels
To explore the association between different CSF amyloid biomarkers (Aβ 42, Aβ 40 , and amyR) and CSF p-tau levels, we performed correlation (see Additional file 3) and linear regression analyses adjusting for age and sex (see Table 2).Given the previous association between amyloid and p-tau markers in the A − T − and A + T − subgroups, we applied a robust locally weighted regression model to trace the trajectories of standardized (z-scores) CSF biomarkers as a function of amyR.We observed that CSF Aβ 42 dramatically declined, reaching − 4.5 z-scores, when amyR levels were still normal, and plateaued when amyR became pathological (− 5 z-scores).Remarkably, CSF Aβ 40 and p-tau also started to increase before the amyR cut-off was reached (2 z-scores for CSF p-tau and 0.8 z-scores for Aβ 40 ), but steeply increased above the pathological threshold of amyR, eventually reaching the highest z-scores (5 z-scores for CSF p-tau and 4.5 z-scores for Aβ 40 ) (Fig. 2).

FDG-PET substudy
Demographics of the participants to the substudy are reported in Additional file 4. Two expert raters (A.M. and A.C.) separately evaluated FDG-PET scans from the 46 patients in the substudy and identified 37 typical AD patterns, four possible AD patterns, and five normal scans, showing good inter-rater agreement (Cohen's k = 0.9).Holding scans with both typical and possible patterns as AD-positive, Cohen's k analysis demonstrated a significant agreement between abnormal FDG-PET and pathological CSF p-tau/Aβ 42 (Cohen's k = 0.73), with a mismatch of 6% between them.
Comparisons between FDG-uptake patterns of each patient group versus CG showed a significant reduction in the temporo-parietal and frontal regions (see Additional file 5).The full factorial design, accounting for inter-group differences, resulted in a significant main effect of groups, and results from the post hoc analysis are shown in Table 4. Compared to the control

Discussion
In the complex AD framework, the discordance between biomarkers, especially those accounting for amyloid status, might encourage both clinicians and researchers to prefer biomarkers with higher specificity, despite the risk of losing sensitivity and likely missing the problem of the different meanings associated with those biomarkers.

Main findings in CSF AD biomarkers and disease burden
CSF Aβ 42 levels did not significantly change across the AD continuum.Conversely, we identified a stepwise increase in CSF Aβ 40 , with lower levels in controls as in CSFAβ 42 + /amyR − , higher levels in CSFAβ 42 + /amyR + , and the highest in A + T + , resembling the CSF variations of p-tau and p-tau/Aβ 42 .
Decreased CSF Aβ 42 levels are a well-established finding in AD and could reflect the aggregation of Aβ 42 in brain tissue [27,28], the formation of semisoluble Aβ 42 oligomers [29], or even the binding of peptides in complexes that mask epitopes targeted by analytical assays [30].Moreover, the hypothetical temporal changes of AD biomarkers presume that the CSF Aβ 42 reduction begins in the pre-clinical stage [31].However, repeated CSF measurements from symptomatic patients showed longitudinal stability of CSF Aβ 42 levels, making this biomarker unsuitable for reflecting the underlying dynamics of amyloid metabolism over time [32].On the other hand, amyR is a better predictor of abnormal cortical amyloid plaque burden in AD [11,12], possibly because the addition of CSF Aβ 40 in the ratio can correct for interindividual variability of amyloid production [33,34].However, another reasonable explanation may lie in a specific role of Aβ 40 in AD. Biochemical studies have shown that Aβ 42 has faster aggregation properties than other Aβ species and that Aβ 40 monomers inhibit its aggregation [35][36][37], by preferentially binding to protofibrillar Aβ 42 [38,39].Moreover, different ratios of Aβ 42 to Aβ 40 can result in different kinetics of amyloid fibrillization and aggregation.Chang and colleagues demonstrated that an equimolar Aβ 40 /Aβ 42 sample generates oligomers with the highest neurotoxic effects on neuritic length, while an increase in the proportion of Aβ 40 stabilizes the fibrillization pathway [40].However, an Aβ 40 -dominant ratio causes changes in calcium dynamics, resulting in a higher elevation of intracellular calcium levels and neuronal apoptosis [41].Thus, we may speculate that CSF levels of Aβ 40 initially increase to quench the fibrillization pathway triggered by Aβ 42 oligomerization; however, beyond a certain threshold, they may have toxic effects on neurons.
When we plotted the changes in CSF biomarkers as a function of amyR, we noticed different trends in CSF biomarkers variations.We observed that Aβ 42 had the steepest decrease before the amyR-positivity cut-off and then encountered a plateau phase.In contrast, CSF p-tau and Aβ 40 started increasing before the pathological threshold of amyR, but this increase was significantly more pronounced when amyR became pathological.This finding is consistent with the notion that cortical Aβ deposition precedes neocortical tau aggregation [34] and with previous studies reporting that the decrease in CSF Aβ 42 / Aβ 40 ratio is followed by a large increase in CSF p-tau also in preclinical AD patients [42].Indeed, we observed that a decrease of Aβ 42 in the CSF is accompanied by an increase of tau phosphorylation but also that the increase of Aβ 40 levels-which significantly affects amyR positivity-is accompanied by a steeper increase in CSF p-tau levels, which has recently been described to boost the spread of tau pathology in the brain [43].
In our cohort, we found that among A + T − patients, CSFAβ 42 + /amyR + showed a higher probability of having pathological levels of p-tau/Aβ 42 , a well-known marker of AD-related burden, than CSFAβ 42 + /amyR − .Considering the differences in CSF levels of Aβ 40, but not Aβ 42 , we might assume that the higher levels of Aβ 40 in CSFAβ 42 + /amyR + could concur with these results.
Unexpectedly, although the CSFAβ 42 + /amyR − condition might be misinterpreted as an A − status if we relied on amyR more than CSF Aβ 42 , in our study, we found that pathological values of p-tau/Aβ 42 can be present also in these patients.This highlights that amyR, commonly held as the best biomarker for detecting amyloid pathology [35], also shows suboptimal negative predictive power, suggesting that the added value of Aβ 40 could lie beyond its Aβ 42 -corrective function, since correcting for interindividual amyloid production variability should improve not only amyR specificity but also sensitivity.
The specific meaning of the altered CSF Aβ 40 levels could be rooted in its significant correlation with the CSF levels of p-tau, which is also present under physiological conditions [44].In our results, we observed a positive linear relationship between Aβ 40 and p-tau in A − T − , CSFAβ 42 + /amyR − , and CSFAβ 42 + /amyR + , but not in A + T + .We cannot exclude that, alongside pathological Aβ 42 levels, there may be a concomitant increase in CSF Aβ 40 (as in CSFAβ 42 + /amyR + patients) associated with a parallel increase in tau phosphorylation.Conversely, in our A + T + group tau pathology seems to proceed independently from amyloid pathology as no association was found between amyloid and tau CSF biomarkers.In line with our results, it has been recently demonstrated that the Aβ-induced increase in CSF p-tau levels might play a key role in initiating tau aggregation and spreading in early AD, while local tau seeding and auto-replication predominate once soluble p-tau concentrations reach a plateau in AD dementia [43].

Main findings in FDG-PET metabolic pattern
To support our hypothesis that CSFAβ 42 + and amyR + hold different meanings in AD pathophysiology, we considered patterns of brain glucose metabolism.Indeed, specific FDG-PET hypometabolic patterns have been demonstrated to predict AD dementia [15][16][17] and may detect additional differences between CSFAβ 42 + / amyR − and CSFAβ 42 + /amyR + .
First, when comparing FDG-PET scans from CSFAβ 42 + / amyR − patients and controls, we found a pattern of cortical hypometabolism encompassing the parietal and frontotemporal regions.Thus, the decrease in CSF Aβ 42 alone may be associated with synaptic dysfunction and reduced brain glucose uptake in areas typically involved in AD [45,46].However, other crucial factors (i.e., such as endothelial dysfunction, neuroinflammation, and astrocytic impairment) could also have influenced synaptic functioning, especially in the very early stages of the disease, before overt amyloid pathology occurs [47][48][49].
In contrast, patients with CSFAβ 42 + /amyR + showed further involvement of the frontal regions, configuring an FDG-PET pattern indistinguishable from A + T + .Previous studies have shown that soluble Aβ 40 , but not Aβ 42 , extracted from the frontal cortex of patients with AD gradually increases along with the progression of Braak scores, suggesting a specific link between Aβ 40 levels and the progression of tau pathology [50].Moreover, a positive correlation between soluble tau levels and hypometabolism in the frontal regions has also been reported [51].Our findings suggest that amyR positivity, which in our cohort is determined by the co-presence of decreased Aβ 42 and increased Aβ 40 levels, is associated with widespread brain hypometabolism, and probably with the spread of tau pathology, even in the absence of elevated tau proteins in the CSF [52].
The lack of detection of elevated levels of CSF p-tau 181 in our cohort does not exclude a possible increase of other p-tau isoforms (e.g., p-tau 217 or p-tau 231 ).Moreover, a threshold-based evaluation of CSF p-tau, as for other fluid biomarkers, can have limitations tied to the risk of false negative results, whose occurrence also encompasses pre-analytical (e.g., collection techniques, handling of samples, storage conditions) and analytical variables (e.g., accuracy in sample processing, assay sensitivity) [53].On the other hand, there may be other yet-to-be elucidated pathophysiological processes underlying a possible mismatch between parenchymal and CSF evidence of tau pathology.Considering all these limitations, a multimodal approach combining imaging and CSF biomarkers could be advisable.
Our results show that among patients classified as A + T − , there is a group with reduced Aβ

Limitations
To our knowledge, this study is the first to directly compare CSFAβ 42 + /amyR − and CSFAβ 42 + /amyR + patients, as well as the first to examine differences and overlap between them, healthy controls, and A + T + in terms of CSF p-tau, AD-related burden, and brain FDG-PET metabolic patterns.However, because of its cross-sectional design, we could not investigate temporal connection between different biomarkers alteration or evaluate whether higher CSF Aβ 40 levels, along with pathological CSF Aβ 42 , may affect the rapidity of cognitive decline.
Furthermore, fluid biomarkers, represented on a continuous scale, are commonly used in a dichotomous way, applying cut-offs whose thresholds become crucial in clinical practice.Such thresholds are not always available for every biomarker measure, and a global effort of standardization is needed to increase accuracy and reproducibility of these findings.Further investigations using amyloid-PET or tau-PET would be useful to verify our results on the prevalence of pathological AD-related burden across patient groups-which we evaluated indirectly via the CSF p-tau/Aβ 42 ratio-and to assess any causal connections between the alteration of Aβ 42 , Aβ 40 , and p-tau in the CSF and parenchymal amyloid and tau deposition.Finally, extending the analysis by comparing blood-based and CSF biomarkers could be useful to confirm these relationships and to assess their extensibility for diagnostic and prognostic purposes.

Conclusions
Fluid biomarkers are nowadays being introduced into clinical routine practice, positing challenges on their correct use for either supporting or excluding AD diagnosis.
Patients with cognitive symptoms and decreased CSF Aβ 42 without pathological amyR nor increased p-tau levels could, under certain circumstances, be essentially considered non-AD, but this conclusion can be hard to reach when clinical presentation and FDG-PET pattern are highly suggestive of AD.Our results suggest that decreased CSF Aβ 42 alone could detect Alzheimer's pathology, since these patients are distinguished from controls by reduced FDG uptake in brain regions typical of AD, and some of them also show pathological measures of AD-related burden.
On the other hand, the increase in CSF Aβ 40 levels, traced by pathological amyR, associates with higher levels of soluble hyperphosphorylated tau, higher AD-related burden, and a widespread pattern of FDG hypometabolism.
Considering the many experimental novel drugs that see Aβ and tau proteins as therapeutic targets, the specific meaning of different amyloid CSF biomarkers should be considered to support a more accurate selection of patients and search for the best time window for treatment.

Fig. 2
Fig. 2 Results of the robust locally weighted regression analysis showing changes of CSF biomarkers as a function of amyR.Values of p-tau, Aβ 42 , and Aβ 40 are expressed as z-score standardized on the average and standard deviation of the A − T − group.The dotted line represents threshold value of pathological amyR (0.06), the dotted arrow indicates the point of change of Aβ 42 from normal to pathological values.CSF, cerebrospinal fluid; amyR, Aβ 42 /Aβ 40 ; A, amyloid status; T, p-tau status) group (CG), CSFAβ 42 + /amyR − patients showed a significant reduction in brain glucose consumption in a wide cluster encompassing the left parietal (BAs 19 and 40), temporal (BA 37), and frontal lobes (BAs 6, 8, and 46) (Fig.3A).With respect to CSFAβ 42 + /amyR − , CSFAβ 42 + /amyR + showed an additional reduction in FDG uptake in the anterior regions, namely the anterior cingulate and frontomedial areas (BAs 32 and 10) (Fig.3B).No differences were found between the hypometabolic patterns of CSFAβ 42 + /amyR + and A + T + patients (p > 0.05).

Table 1
Demographic and clinical characteristics across AT groups Data are presented as mean ± standard deviation or percentages, when applicable.CSF cerebrospinal fluid, Qalb albumin quotient, MMSE Mini-Mental State Examination, p p-value.MMSE score of ≥ 24 was used as a cut-off to separate MCI from dementia patients.Percentages indicate subjects showing the presence of each variable within the group.Bold values represent p-value < 0.05

Table 3
Results of multivariate regression analysis exploring the odds of having pathological values of p-tau/Aβ 42 in A + T − patients

Table 4
Numerical results of SPM comparisons of FDG uptake in CG vs. CSFAβ 42 + /amyR − , CSFAβ 42 + /amyR − vs. CSFAβ 42 + /amyR + , and CSFAβ 42 + /amyR + vs.A + T +In the "cluster level" section (left), the number of voxels, corrected p-value of significance, and cortical region where the voxel is found are all reported for each significant cluster.In the "voxel level" section, all the coordinates of the correlation sites (with the Z-score of the maximum correlation point), the corresponding cortical region and BA are reported for each significant cluster (CG, control group; Inf., infinite; L, left; R, right; BA, Brodmann area).When the maximum correlation is achieved outside the gray matter, the nearest gray matter (within a range of 5 mm) is indicated by the corresponding BA