Combining plasma Aβ and p-tau217 improves detection of brain amyloid in non-demented elderly

Alzheimer's disease (AD) is the most common neurodegenerative disorder and the leading cause of dementia worldwide, threatening aging societies with a vastly increasing number of patients with dementia, and its economic and social burden. Two disease-modifying therapies (DMTs) targeting amyloid-β (Aβ) pathology, aducanumab and lecanemab, have recently been approved by the FDA for use in the early symptomatic stage of AD [1 –4]. Another anti-Aβ drug, donanemab, has met the primary and secondary cognitive endpoints in its phase 3 clinical trial [5]. While the slowing of cognitive decline in response to these therapies was modest, results from the donanemab and lecanemab trials [6, 7] suggest that Aβ-targeting DMTs may be more effective in the earliest stages of AD[3, 4, 6, 7]. This will most likely lead to a shift in the target population of future clinical trials of DMTs to preclinical and prodromal AD. However, recruitment of participants in the earliest stages of AD is challenging due to the low prevalence of preclinical AD in cognitively normal individuals and those with subjective cognitive decline [8] and the invasiveness and high cost of the current gold standard markers, i.e., amyloid positron emission tomography (PET) scans and cerebrospinal fluid (CSF) biomarkers. The use of emerging blood-based biomarkers in the screening of potential trial participants has been highlighted as an efficient approach to overcome these limitations [9, 10]. Although promising results on plasma biomarkers, e.g., Aβ(1–42) (Aβ42) reductions [11 –13] or increases in phosphorylated tau [14] have recently been reported, the accuracy of the combination of plasma biomarkers in predicting amyloid status in individuals at the preclinical or prodromal stages of AD has not been fully investigated. In addition, the effect of ethnic differences on the predictive power of plasma biomarkers has not been well characterized [15, 16]. In this study, we demonstrated the very high performance of the combination of plasma Aβ and p-tau217 to detect brain Aβ-PET positivity in people with early-stage AD in the Japanese J-TRC cohort, which was replicated in the second Caucasian BioFINDER cohort.

In this study, we have shown that the plasma Aβ42/40 ratio determined by IP/MS and plasma p-tau217 measured by MSD immunoassay are biomarkers that predict brain Aβ PET positivity in the Japanese population of non-demented individuals, and that a combination of these Aβ42 and p-tau217 markers showed unprecedented high discriminative values with an AUC of ~ 0.93, which was reproduced in the European BioFINDER cohort. Emerging evidence supports the importance of newer blood-based markers in the detection of cerebral Aβ pathology [10, 22]. In a head-to-head comparison study of several plasma Aβ assays, MS-based plasma Aβ biomarkers were reported to be generally superior to immunoassays in detecting abnormal brain amyloid status: IP-MS method developed by Washington University showed the highest AUC of 0.852 for CSF Aβ42/40 status in cognitively unimpaired and MCI subjects, which was improved to 0.882 by the addition of APOE genotype [22]. Here we showed that the plasma Aβ42/40 ratio, as determined by the Shimadzu-developed IP-MS assay, predicted brain Aβ PET positivity in the J-TRC cohort, which enrolled non-demented elderly individuals by consecutive recruitment through web-based participation and local cohort or memory clinic, thus reflecting the characteristics of elderly individuals in the general population. Plasma Aβ42/40 measures have been shown to be highly discriminative of CSF Aβ42/40 and Aβ PET status, as well as predictive of cognitive decline and progression from cognitively unimpaired to MCI and from MCI to AD [11, 12, 27]. Changes in the plasma Aβ42/40 ratio precede elevated amyloid levels detected by PET scans, similar to those in CSF [9, 28]. Notably, plasma Aβ42/40 showed moderate accuracy in both the cognitively unimpaired (CDR 0) and the MCI (CDR 0.5) subjects, while the accuracy of the plasma p-tau217 was higher, and in the CDR 0.5 group, p-tau217 showed high accuracy (AUC 0.925). It has been well documented that high levels of plasma p-tau, especially p-tau217, are associated with abnormal Aβ PET and CSF Aβ42/40 in different stages of AD [9, 21, 28, 29]. Elevated plasma p-tau levels have been shown to be highly specific for brain amyloid deposition, allowing the differentiation of AD from non-AD dementia [21, 23, 30 –32]. Furthermore, a recent study suggests that plasma p-tau has a strong surrogacy in preclinical and prodromal AD compared to Aβ42/40 ratio or p-tau231 [33]. The superiority of p-tau217 over Aβ42/40 ratio, especially in the prodromal population, is confirmed in BioFINDER. Furthermore, our study suggested an interesting difference in the predictive ability between Aβ42/40 ratio and p-tau217. As shown in Table 3, Aβ42/40 ratio showed relatively higher sensitivity (0.904) and lower specificity (0.733) in the CDR 0 group, while the results were opposite in the CDR 0.5 group (0.743, 0.875, respectively). P-tau217 showed the opposite result (sensitivity 0.761, specificity 0.896 in the CDR 0 group while 0.897 and 0.826, respectively, in the CDR 0.5 group). These results indicate that when blood biomarker is used as a pre-screen for amyloid PET, p-tau217 reduces false positive compared to Aβ42/40 ratio in the CDR 0 group, while Aβ42/40 ratio may be better compared to p-tau217 in the CDR 0.5 group. These results are not replicated in the BioFINDER cohort, where p-tau217 showed higher specificity compared to Aβ42/40 ratio regardless of group. This difference may be due to the difference in disease progression between CDR 0.5 in J-TRC and MCI in BioFINDER, as shown by the plasma p-tau217 level being higher in MCI in BioFINDER (0.28 ± 0.15) compared to CDR 0.5 group in J-TRC (0.24 ± 0.18).

Our results also suggest an intriguing difference in the optimal combination of plasma Aβ42, Aβ42/40, p-tau217 and clinical information in detecting abnormal Aβ PET between the CDR 0 and 0.5 populations. The combination of p-tau217 and Aβ42/40 ratio showed the best performance in the CDR 0 population, while the p-tau217/Aβ42 ratio performed best in the CDR 0.5 group. These results were reproduced in BioFINDER. One interpretation of these findings would be that the plasma Aβ42/40 ratio and p-tau217 gradually change with the progression of brain amyloid accumulation before the threshold of Aβ-PET positivity is reached; the optimal combination of Aβ and p-tau markers may change with disease progression. The benefit of combining plasma Aβ and p-tau biomarkers has not been extensively studied. The superiority of the combination of high performance plasma p-tau217 and Aβ42/40 assays to identify brain Aβ positivity, predict the presence of AD neuropathologic changes [34], Aβ PET centiloid metric [35], and future development of AD dementia has only been reported [36]. Our study demonstrated that the combination of p-tau217 and Aβ42/40 ratio, and these with age, sex, and APOE genotype, is a useful tool for predicting Aβ-PET positivity in the cognitively unimpaired or preclinical population. The contribution of each variable to the modeling is shown by a nomogram in Supplementary Figs. 1 and 2.

In many of the previous studies, participants recruited from regional or clinical cohorts included individuals at high-risk for AD with a higher prevalence of the APOE ε4 allele (e.g., 45.2% in AHEAD 3–45 study [35], 47.5% in the BioFINDER study [23], 53.9% in the ALFA + cohort [32] and 47.8% in the Wisconsin Registry for Alzheimer's Prevention cohort [31]), whereas the APOE ε4 positivity in the J-TRC cohort was 20.2%, which is close to that of the general population in Japan and Asia [37]. Our study used Aβ-PET positivity as the standard of truth, which was determined by visual interpretation in J-TRC and rated with quantitative measures in BioFINDER. While quantitative methods are popular in Europe, visual assessment is the standard in Japan, as indicated by Japanese guidelines and regulations. As a result, there is variation in amyloid PET interpretation criteria depending on the research protocol. This study examined the reproducibility in two cohorts, each with its own protocol including Aβ-PET rating method. In general, visual and quantitative ratings are known to provide comparable results, allowing for error in borderline cases and taking into account possible differences in case of localized uptake. In this study, the inter-rater variability of the visual reading of the J-TRC was low; the two independent raters agreed in 94% of cases. Also, quantitative measures are not free from variation due to bias, e.g., selection of a software program and a cutoff level. Therefore, the difference in PET rating method between the two cohorts would not substantially affect the conclusions of our study.

Our results may be relevant to the use of plasma biomarkers for prescreening in clinical trials of preclinical AD populations in the real world. Our study also addresses the potential impact of ethnic factors on the usability of plasma biomarkers [15, 38]. Our results showed high performance of the combination of plasma Aβ42, Aβ42/40 ratio, and p-tau217 in non-demented elderly individuals in Asian-Japanese as well as Western populations. Thus, these biomarkers could greatly facilitate the prescreening of participants in global preclinical AD trials that require the enrollment of participants from diverse ethnicities. In addition, these blood-based biomarkers may play an important role in the early detection of individuals at high risk of developing AD and for the early and appropriate diagnosis of cognitive decline or dementia due to AD.

This study has several limitations. Other promising plasma biomarkers such as other phosphorylated tau species (e.g., p-tau231), phosphorylated/non-phosphorylated tau ratio, GFAP, NFL should also be investigated. Our study population of CDR 0.5 in J-TRC and MCI in BioFINDER is relatively small. The rate of APOE ε4 allele carriers is not the same in the two cohorts. Thus, the generalizability of our findings, e.g., the difference of optimal combination for cognitively impaired or MCI population in the real-world setting, should be verified in future studies.

This study demonstrated a high accuracy of the combination of plasma Aβ markers and p-tau217 to detect Aβ-PET positivity in preclinical and prodromal AD in the Japanese trial-ready cohort, which was replicated in the Swedish BioFINDER cohort. These results provide us with optimal indices to identify potential participants and minimize the financial and physical burden of clinical trials of DMTs in the very early stages of AD.