The clinical application of optimized AT(N) classification in Alzheimer’s clinical syndrome (ACS) and non-ACS conditions

We aimed to assess the utility of AT(N) classification in clinical practice. We measured the cerebrospinal fluid levels of amyloid-β (Aβ) 42, Aβ40, phosphorylated tau, total tau, and neurofilament light chain (NfL) in samples from 230 patients with Alzheimer's clinical syndrome (ACS) and 328 patients with non-ACS. The concordance of two A-markers (i.e., Aβ42 alone and the Aβ42/Aβ40 ratio) was not significantly different between the ACS (87.4%) and non-ACS (74.1%) groups. However, the frequency of discordant cases with AAβ42-alone+/AAβ-ratio- was significantly higher in the non-ACS (23.8%) than in the ACS group (7.4%). The concordance of two N-markers (i.e., total tau and NfL) was 40.4% in the ACS group and 24.4% in the non-ACS group. In the ACS samples, the frequency of biological Alzheimer's disease (i.e., A+T+) in Ntau+ cases was 95% while that in NNfL+ cases was 65%. Reflecting Aβ deposition and neurodegeneration more accurately, we recommend the use of AT(N) classification defined by cerebrospinal fluid AAβ-ratioTNNfL in clinical practice.


Introduction
The global prevalence of dementia is predicted to increase exponentially over time as a result of the lengthening of human lifespan (GBD 2019Dementia Forecasting Collaborators, 2022. Alzheimer's disease (AD) is the most common cause of dementia, and is pathologically characterized by β-amyloid (Aβ) deposition and fibrillar phosphorylated tau (pTau) accumulation, followed by neurodegeneration and synapse loss in brain (DeTure and Dickson, 2019). After decades of research and development of disease-modifying therapies for AD, the anti-Aβ antibody, aducanumab, was recently approved by the US Food and Drug Administration. Other types of anti-Aβ antibodies are expected to become available in clinical practice shortly. However, the intervention using disease-modifying therapies for AD requires accurate in vivo detection of the underlying pathological changes.
Because of the proven utility of cerebrospinal fluid (CSF) biomarkers in the detection of pathological changes that occur in AD, they are used as defining markers in the research framework criteria . The use of AT(N) classification, using CSF Aβ42 or the Aβ42/Aβ40 ratio as an Aβ (A) marker, pTau as a pathological tau (T) marker, and total tau (tTau) as a neurodegeneration (N) marker has been proposed in the research framework . CSF neurofilament light chain (NfL) has also been found useful as an Nmarker (Khalil et al., 2018). Compared with Aβ42 alone, the Aβ42/ Aβ40 ratio has been shown to result in more accurate differential diagnoses of AD, improved prediction of progression from mild cognitive impairment (MCI) to AD dementia, and higher concordance with amyloid positron emission tomography (PET) findings . Regarding CSF N-markers, tTau and NfL are not always well correlated, suggesting that these markers may reflect different aspects of neurodegeneration (Kasuga et al., 2022;Mattsson et al., 2016;Van Hulle et al., 2021). The term Alzheimer's clinical syndrome (ACS) is clinically defined by amnestic syndrome or classic variants and applied to MCI and dementia . There are several previous studies comparing the diagnostic utility of two A-markers (Amft et al., 2022;Delaby et al., 2022;Dumurgier et al., 2015;Janelidze et al., 2016;Spies et al., 2010) or two N-markers (Cousins et al., 2021), separately. However, there has been no previous research that addresses the performance of these two A-and two N-markers in the same patients in ACS and non-ACS groups by a multi-center study.
In research cohorts at all points on the AD spectrum from cognitively unimpaired (CU) through MCI to AD dementia, the usefulness of the AT(N) classification for evaluating AD pathology has been validated both in Western (Eckerstrom et al., 2021;Ekman et al., 2018;Grontvedt et al., 2020;Kern et al., 2018;Mattsson-Carlgren et al., 2020;Soldan et al., 2019) and Asian countries (Hu et al., 2022;Kasuga et al., 2022;Lee et al., 2020). In these patients with AD dementia, the prevalence of biological AD positive for both A-and Tmarkers was found to be 57%-82%. This suggests that at least 20%-40% of such patients are likely to have conditions that mimic AD; these are clinically similar to AD but lack AD pathology. Although these CSF biomarkers have been widely applied in clinical practice (Delaby et al., 2021;Skillback et al., 2015), there have been few reports evaluating the utility of AT(N) classification in clinical practice (Carandini et al., 2019;Ye et al., 2021).
In this study, we aimed to clarify: (1) the difference of the performance of the two A-markers and the two N-markers; (2) the prevalence of biological AD; and (3) the frequency of AT(N) categories defined by the two A-and two N-markers among the ACS and non-ACS patients in a Japanese clinical cohort. This information will be helpful in the implementation of these markers in clinical settings.

Participants
A total of 673 CSF samples were obtained for diagnostic purposes at Niigata University and related facilities between October 2013 and June 2022. Of these, 19 follow-up samples were taken after the first visit and 14 samples without clinical diagnoses were excluded (Fig. 1) Table S1 without information on CSF biomarkers or amyloid/tau PET. Patients with CAA were excluded from this study because CAA shares common Aβ pathology with AD. Others represented a group of heterogeneous neurological disorders and were also excluded because including these patients in the non-ACS group may complicate the interpretation of the results. As a result, 558 subjects were included for biomarker analyses in this study. Patients with probable AD, possible AD, and MCI due to AD were grouped as ACS and the remaining patients as non-ACS (Fig. 1).
The information on age at lumbar puncture, sex, and Mini-Mental State Examination (MMSE) scores were obtained from case report forms. MMSE scores were missing in 39 patients with ACS and 120 patients with non-ACS. We performed a subgroup analysis of the ACS group stratified by age at lumbar puncture (< 65 years [n = 59] or ≤ 65 years [n = 171]).
This study was conducted in accordance with the Declaration of Helsinki and approved by the Ethics Committee of Niigata University (2019-0239). All participants or their representatives provided written informed consent to participation.

CSF collection and analysis
CSF samples were collected by lumbar punctures at each institution and sent to Niigata University. Polypropylene collection tubes were used as much as possible but were not available at all sites. In both the ACS group and non-ACS group, results of A-markers and N-markers concordance/discordance in the subgroup with only the use of a polypropylene tube were not different from the whole group (data not shown). CSF was aliquoted at a volume of 0.5 mL and stored at -80 ℃ until the assay.
Up until March 2019, the CSF concentrations of Aβ42, tau phosphorylated at threonine 181 (pTau), and tTau were measured using an AlzBio3 kit (Fujirebio, Ghent, Belgium). After March 2019, CSF concentrations of pTau and tTau were measured using INNOTEST PHOSPHO-TAU (181P) and hTAU Ag (Fujirebio, Ghent, Belgium). This was due to the suspended supply of AlzBio3 kits. The CSF concentrations of Aβ42 and Aβ40 were measured using V-PLEX Aβ Peptide Panel 1 (6E10) (Meso Scale Discovery, Rockville, MD, USA), and the Aβ42/Aβ40 ratio was calculated. The Aβ42 values obtained using V-PLEX, and the pTau and tTau values obtained using INNOT-EST were converted to the measurement values used by the AlzBio3 kit based on data from bridging samples of available CSF biomarker data using both methods. (Aβ42: n = 753, Spearman's r = 0.910; pTau: n = 49, Spearman's r = 0.780; tTau: n = 51, Spearman's r = 0.889). CSF concentrations of NfL were measured using R-PLEX Human Neurofilament L Antibody Set (Meso Scale Discovery, Rockville, MD, USA). The absolute NfL concentrations differed between our cohort and previous studies that used the UmanDiagnostics kit, and this is likely due to the assays using different antibodies and calibrators (Bridel et al., 2019;Kasuga et al., 2022). All analyses were conducted in duplicate by experienced laboratory personnel blinded to the clinical diagnosis. The intra-assay and inter-assay coefficients of variation were < 20% for all assays. The laboratory at Niigata University participates in the Alzheimer's Association external quality control program for CSF biomarkers (Mattsson et al., 2013).

Cutoff values and classification of biomarker status
In a previous study, we determined cutoff values for Aβ42, pTau, tTau, and NfL based on samples from the Japanese Alzheimer's Disease Neuroimaging Initiative (J-ADNI) (Kasuga et al., 2022). The cutoff value of the Aβ42/Aβ40 ratio was determined by Gaussian mixture models (GMM) based on 177 samples from the J-ADNI ( Fig.  S1 and Table S2). The cutoff value for Aβ42/Aβ40 ratio obtained using GMM was slightly higher than values calculated using the Youden index based on clinical status (CU vs. AD dementia), PET status (Aβ PET− vs. Aβ PET+), and a combination of these (CU with Aβ PET− vs. AD dementia with Aβ PET+). We determined cutoff values by GMM for all biomarkers except for NfL because the GMM method is unbiased to detect AD pathology and more sensitive for detecting biological changes (Bertens et al., 2017;De Meyer et al., 2010). The distribution of NfL was unimodal, which is unsuitable for GMM. To obtain a cutoff value of NfL, we compared Aβ PET−, CU group with Aβ PET+, AD dementia group in our previous study using an independent cohort (Kasuga et al., 2022). The cutoff values used in this study were CSF Aβ42 < 359.6 pg/mL as A Aβ42-alone +, Aβ42/Aβ40 ratio < 0.072 as A Aβ-ratio +, pTau > 30.6 pg/mL as T+, tTau > 105.3 pg/ mL as N tau +, and NfL > 2650 pg/mL as N NfL +. The A+T+ profile is referred to as biological AD according to the research framework . To clarify the biomarker changes, CSF biomarkers in ACS and non-ACS groups were compared with those of CU controls (n = 53, median age 67, 52.8% female) from our previous study (Kasuga et al., 2022) (Fig. S2).
As an emerging T-marker, CSF pTau/Aβ40 ratio has been reported to be an excellent measure of tauopathy in AD (Guo et al., 2020). We have compared pTau with pTau/Aβ40 as T-marker in this study (Fig. S3). Because the bimodality of pTau/Aβ40 distribution was not clear, we compared the CU group with Aβ PET− and the AD dementia group with Aβ PET+ in the independent cohort (J-ADNI) to obtain a cutoff value based on the Youden index. Aβ42/Aβ40 ratio positivity in cases with pTau/Aβ40+ was comparable with cases with pTau+ in both ACS (90% vs. 93%) and non-ACS group (65% vs. 66%) (Fig. S3). Because pTau/Aβ40 had comparable diagnostic performance to pTau in our cohort. In this study, we have chosen to use CSF pTau as Tmarker.

Statistical analyses
Statistical analyses were performed using GraphPad Prism software (GraphPad Software Inc., La Jolla, CA, USA) and the R software (http://www.r-project.org/). Group comparisons were performed using Mann-Whitney U tests for continuous data such as age and MMSE score. Group comparisons for categorical data were performed using χ 2 tests or Fisher's exact tests. Group differences in CSF biomarker levels were assessed with logistic regression models that included age and sex as covariates. Correlations between the two datasets were identified using Spearman's rank-correlation coefficient. The percentage of concordance/discordance of two A-markers, or two N-markers, and prevalence estimates for the AT categories were calculated with 95% confidence intervals (CI) generated using bootstrap resampling (n = 1000).

Participant demographics
A total of 558 CSF samples from 230 patients with ACS and 328 patients with non-ACS were analyzed. The patients' demographic characteristics and CSF biomarker levels are shown in Table 1 and Fig. S2. Median age at lumbar puncture was not different between the ACS group and the non-ACS group. The ACS group contained more females than the non-ACS group. The MMSE scores of the ACS group were lower than those of the non-ACS group. The levels of CSF biomarkers were compared between the ACS and non-ACS groups after adjustment for age and sex. Aβ42 levels were comparable, but the Aβ42/Aβ40 ratio was significantly lower in the ACS than in the non-ACS group. The levels of pTau and tTau were significantly higher in the ACS group. The levels of NfL were numerically higher in the non-ACS group, although the difference did not reach statistical significance. The demographic characteristics and CSF biomarker levels of the various clinical diagnostic groups are shown in Table S3 and Fig. S4.
We next evaluated the frequency of pTau positivity in discordant cases for A-markers ( Fig. 2C and D). The pTau positive rate was 71%-83% in A Aβ42-alone −/A Aβ-ratio +, and 6%-8% in A Aβ42-alone + /A Aβ-ratio − in the ACS and non-ACS groups, respectively. This suggests that A Aβ42-alone −/A Aβ-ratio + reflects AD pathology, but the A Aβ42-alone + /A Aβ-ratio − does not do in discordant cases. Additionally, in discordant cases for A-markers in the ACS group, age, and MMSE score were not different between A Aβ42-alone −/A Aβ-ratio + and A Aβ42-alone +/A Aβ-ratio − (data not shown). This indicates that the lower pTau positive rate in A Aβ42-alone +/A Aβ-ratio − is not due to earlier disease detection than in A Aβ42-alone −/A Aβ-ratio +.
We next compared the N-marker positivity of CSF tTau and NfL. In the ACS group, only 27.4% of cases were positive for tTau, which was significantly less than the 82.6% of NfL positive cases (p < 0.0001) (Fig. 3C). Similarly, in the non-ACS group, only 12.2% of cases were positive for tTau, which was significantly less than the 86.6% positive for NfL (p < 0.0001) (Fig. 3D). In biological AD (i.e., A Aβ-ratio +/T+) patients in the ACS group, the proportion of N tau + (40.5%) was significantly lower than that of N NfL + (83.1%). This suggests that NfL may be a more sensitive N-marker for neurodegeneration in AD pathology than tTau. Further, we compared the proportion of biological AD in N-marker positive ACS cases. Around 95.2% of N tau + cases showed A+T+, while only 64.7% of N NfL + cases showed A+T+ (p < 0.0001). This suggests that most N tau + ACS cases have biological AD change, while only 65% of N NfL + ACS cases were considered to have biological AD change. Elevation of NfL is partly caused by pathological conditions other than AD in ACS. The concordance of N-markers in each clinically diagnosed group is shown in Table S3 and Fig. S6.
On the AD continuum, the A+T− profile is referred to as AD pathologic change . When the Aβ42/Aβ40 ratio was used as the A-marker, there was a lower frequency of AD pathologic change (i.e., A+T−) and a higher frequency of A−T− cases than when Aβ42 alone was used as the A-marker, especially in non-ACS patients ( Fig. 4C and F).

AT(N) classification using different A-and N-markers
Finally, we compared the AT(N) classifications defined by the two different A-markers (Aβ42 alone vs. the Aβ42/Aβ40 ratio) and the two different N-markers (tTau vs. NfL) (Fig. 5). In the ACS group, there was little difference between the classifications defined by the two A-markers. The proportion of N+ patients increased when NfL was used as the N-marker. In the non-ACS group, when the Aβ42/ Aβ40 ratio was used as the A-marker, the proportion of A−T− cases was markedly higher and the proportion of A+T− was markedly lower than when Aβ42 alone was used as the A-marker. The proportion of N+ was higher when NfL was used as the N-marker than when tTau was used. This was to a similar degree to that seen with the ACS samples.

Discussion
In this study, we found that the A-markers Aβ42 and the Aβ42/ Aβ40 ratio to be highly consistent with one another in ACS CSF samples: however, in non-ACS samples, there was a substantial number of discordant cases (23.8%) showing only A Aβ42-alone +. Most non-ACS patients with A Aβ42-alone +/A Aβ-ratio − were negative for pTau (92.3%), suggesting that the reduced Aβ42 in these patients was independent of Aβ deposition in the brain. In both the ACS and non-ACS groups, there was a higher prevalence of N NfL + than N tau +. Patients with N NfL + showed a significantly lower frequency of biological AD (A+T+) than those with N tau +, indicating that NfL levels can be elevated in non-AD pathological conditions as well as AD. In our ACS group, about 60% of cases were A+T+. Therefore, we may infer that approximately 40% of ACS cases are AD mimics that lack Aβ or tau, or both pathologies. In contrast, approximately 25% of the non-ACS cases were A+T+, suggesting that these patients either exhibited atypical AD phenotypes with primary AD pathological changes, or that primary non-AD pathological changes occurred alongside concomitant of AD pathology. Finally, we found that the frequency of different AT(N) profiles differed significantly depending on the choice of N-markers in ACS and on the choice of both A-and N-markers in non-ACS patients.
Although both CSF Aβ42 alone and the Aβ42/Aβ40 ratio are defined as the A-marker reflecting amyloid pathology in the research framework , previous research has shown the Aβ42/Aβ40 ratio to have better diagnostic performance (Amft et al., 2022;Delaby et al., 2022;Dumurgier et al., 2015;Hansson et al., 2019;Janelidze et al., 2016;Spies et al., 2010). Depleted levels of Aβ42 in CSF samples can have causes other than Aβ deposition in the brain, including reduced Aβ generation due to neuronal loss, and faulty Aβ clearance mechanisms in the brain (Lewczuk et al., 2018). Because the Aβ42/Aβ40 ratio in non-AD conditions is less influenced by these alternative mechanisms, reductions in the Aβ42/Aβ40 ratio more accurately reflect Aβ pathology. Delaby et al. found that subjects with A Aβ42-alone −/A Aβ-ratio + have higher levels of CSF pTau+ than those with A Aβ42-alone +/A Aβ-ratio − (Delaby et al., 2022). This is consistent with our results in the present study, which showed A Aβ42-alone −/A Aβ-ratio + patients had a higher rate of pTau positivity compared to A Aβ42-alone +/A Aβ-ratio − patients in both the ACS (83.3% vs. 5.9%) and non-ACS (71.4% vs. 7.7%) groups. In addition, the frequency of discordant cases for the A Aβ42-alone + /A Aβ-ratio − classification was higher in the non-ACS than in the ACS groups (23.8% vs. 7.4%). Because a decrease of CSF Aβ40 may occur by reduced Aβ production or Aβ clearance from the brain, Aβ42/ Aβ40 ratio could differentiate Aβ deposition from these situations (Lewczuk et al., 2018). Taking iNPH as an example of the non-ACS group, 34.1% were classified as A Aβ42-alone +/A Aβ-ratio −. It has been suggested that the decreased CSF Aβ42 levels seen in iNPH are unrelated to Aβ deposition in the brain (Graff-Radford, 2014). Our findings suggest that the Aβ42/Aβ40 ratio may be a better marker than Aβ42 alone for the detection of AD pathology in both ACS and non-ACS patients. When applying the AT(N) classifications to non-ACS patients, the use of Aβ42 alone as an A-marker may increase the false positive rate. Thus, we recommend the use of the Aβ42/Aβ40 ratio as the preferred A-marker, especially in non-ACS patients.
The research framework has defined CSF tTau as an N-marker . However, we and others have observed a high concordance between tTau and pTau (Kasuga et al., 2022;Van Hulle et al., 2021). Hence, CSF tTau is unlikely to be a fully independent marker of neurodegeneration. CSF NfL is elevated in various neurological disorders and its usefulness as an N-marker has been increasingly recognized (Khalil et al., 2018). We previously reported the correlation between tTau and NfL was weaker than that between tTau and pTau, and different positive rates of CSF tTau and NfL as an N-marker in the research cohort covering the AD spectrum (Kasuga et al., 2022). In this study, we used clinical samples from ACS and non-ACS patients. Because tTau strongly correlates with pTau, tTau may increase by mild neurodegeneration in the ACS group (pTau positive rate > 70%). In contrast, in the non-ACS group (pTau positive rate < 40%), tTau could be less affected by pTau elevation and may largely reflect neurodegeneration. This may explain the stronger correlation between tTau and NfL in the non-ACS group than in the ACS group. The mean NfL levels in N tau + ACS group were much less than in N tau + non-ACS group (6344.2 pg/mL vs. 55,710.5 pg/mL). This supports that tTau and NfL may reflect different aspects of neurodegeneration. In both the ACS and non-ACS groups, the use of NfL as an N-marker produced a higher incidence of N+ than the use of tTau as an N-marker. In A+T+ patients in the ACS group, the proportion of N NfL + was significantly higher than that of N tau +. It suggests that NfL would be more sensitive than tTau for detecting neurodegeneration in AD. It should be noted that the tTau positive rate in our ACS group is lower than in previous reports. In cases with ACS (i.e., MCI to AD dementia), the tTau positive rate was 34%-65% in the US-ADNI (Ekman et al., 2018), and 31%-58% in a clinical setting cohort (Carandini et al., 2019), which are higher than 27.4% in our study. Skillback et al. reported that CSF tTau levels were highest in earlyonset AD, followed by late-onset AD, and mixed AD and vascular groups . The ages of the ACS group (median 76 years) were older than those of the J-ADNI cohort (median 71 years), which was used to determine the cutoff of CSF tTau. Thus, the ACS group may have more co-pathology. This factor may explain the low tTau positivity rate in the ACS group in this study. When we used the cutoff value at 88.8 pg/mL for tTau, which separates AD dementia with positive amyloid PET from CU with negative amyloid PET (Kasuga et al., 2022), tTau positive rate in ACS increased to 45.7%. This tTau positive rate is comparable to previous reports (Carandini et al., 2019;Ekman et al., 2018), but still lower than NfL positive rate (83%) in ACS. N NfL + patients had a lower proportion of A+T+ than N tau + patients in both the ACS and non-ACS groups, suggesting that elevated NfL reflects pathologies other than AD. NfL appears to be a better N-marker than tTau for general neurological disorders. Comparison with other N-markers such as MRI is needed to clarify to what extent tTau reflects neurodegeneration in the AD continuum.
AD is defined in vivo by the positivity of both A-and T-markers in the research framework . Herein, we evaluated the effects of two different A-markers on AT classification. In the ACS group in our study, the frequencies of A−T− and A−T+ appear to be comparable with previous reports (Carandini et al., 2019;Ekman et al., 2018;Grontvedt et al., 2020;Mattsson-Carlgren et al., 2020). In our clinical cohort, the frequency of patients with AD pathology (i.e., A+T+) was around 60% for ACS and 25% for non-ACS. More specifically, the frequencies of A+T+ using the Aβ42/Aβ40 ratio as an Amarker were 67.6% in probable AD, 34.6% in possible AD, 69.5% in MCI due to AD, 30.4% in VCI, 35.3% of LBD, 10% in MSA, 25.7% in CBS, 10.3% in PSP, 28.3% of FTLD, 18.7% in iNPH, and 27.9% in the unclassified cognitive impairment subgroup. These A+T+ proportions are comparable to those reported by autopsy-confirmed studies (Chare et al., 2014;Irwin et al., 2017;Jellinger, 2013;Leinonen et al., 2010;Monsell et al., 2015;Robinson et al., 2018;Tsuboi et al., 2003). Further, these A+T+ proportions in both the ACS and the non-ACS group are also comparable to previous reports using CSF Aβ42 and pTau in clinical settings (Carandini et al., 2019;Grontvedt et al., 2020;Lee et al., 2020;Rosen et al., 2015;Skillback et al., 2015). Some of the non-ACS patients with A+T+ may have atypical phenotypes such as CBS and FTD/PPA due to primary AD pathology, or the presence of concomitant AD pathology and other neuropathological changes . Thus, our results indicate that CSF biomarkers are helpful in the detection and exclusion of AD pathology in clinical practice.
Finally, among the AT(N) profiles, A+T+(N)+ was the highest in ACS patients when using the A Aβ-ratio TN NfL classification. In non-ACS patients, non-AD pathological changes (A−T+[N]− or A−T−[N]+ or A−T+[N]+) were the most frequent in the A Aβ-ratio TN NfL classification. Because the Aβ42/Aβ40 ratio and NfL more accurately reflected Aβ deposition and neurodegeneration, respectively, in the non-ACS group, application of the A Aβ-ratio TN NfL classification appears to be optimal in clinical practice.
This study had several limitations. First, the clinical diagnoses were not confirmed by autopsy. To explicitly evaluate the diagnostic value of specific combinations of CSF biomarkers, a comparison with pathological findings as a gold standard is needed. In particular, the impact of co-pathology on biomarker changes should be verified by autopsy. Second, we could not compare our CSF biomarker findings with those from PET imaging in this study, because small numbers of patients underwent PET scans. The correlation analysis between CSF biomarkers and PET images warrants further study. Third, there was limited information on APOE genotypes in this study. APOE genotypes can affect the frequency of A+, especially in non-ACS patients. Fourth, caution is required for the absolute values of NfL, particularly in the cutoff value used in this study. We measured levels of NfL using the electrochemiluminescence system (Meso Scale Discovery) (Kasuga et al., 2022), which is different from previous reports (Bridel et al., 2019). Fifth, we could not match the cognitive level between ACS and non-ACS groups. MMSE score in the ACS group was significantly lower than that of the non-ACS group. Finally, there may have been recruitment bias because patients with difficult differential diagnoses are more likely to be tested for CSF biomarkers in clinical settings. Caution should be taken in the generalizability of some of our findings, such as the frequency of A+T+.

Conclusions
We have demonstrated the usefulness of CSF biomarkers for the detection and exclusion of AD pathology in ACS and non-ACS patients in a Japanese clinical cohort. For in vivo estimation of pathological changes in clinical practice, the use of CSF A Aβ-ratio TN NfL is recommended as this accurately reflects Aβ deposition and neurodegeneration.

Ethics approval and consent to participate
This study was conducted in accordance with the Declaration of Helsinki and after approval of the Ethics Committee of Niigata University. All subjects or legal representatives provided written informed consent to use their medical data and biomaterials for scientific research.

Author contributions
KK contributed to the concept of the study, analysis of the data, and wrote the manuscript. TTs contributed to the acquisition of the data and analysis of the data. MK contributed to the analysis of the data. TTa, KW, SS, HY, YK, RY, HM, YA, KO, OO, TIw, and J-ADNI contributed to the acquisition of the data. AM contributed to the interpretation of the data. TIk contributed to the concept of the study, drafting the manuscript, and critical revision of the manuscript and is responsible for the overall content as guarantor. J-ADNI contributed to the acquisition of the data. All authors reviewed the manuscript. The authors read and approved the final manuscript.

Funding
This study was supported by Japan Agency for Medical Research and Development (AMED) under grant numbers JP22dm0207073, JP22dk0207057, JP22dk0207059, and JP22wm0525019.

Availability of data and materials
The datasets used and/or analyzed during the current study are available from the corresponding author upon reasonable request.

Disclosure statement
There are no competing interests.