Accurate blood tests for Alzheimer’s disease (AD) are critical tools that have the potential to revolutionize dementia research, clinical trials and clinical care. Models combining blood-based biomarkers that represent multiple aspects of AD brain pathology with key individual level factors may improve prediction of AD dementia.
For 80 years after Alois Alzheimer identified the hallmark amyloid plaques and tau tangles of his eponymous dementing illness, Alzheimer’s disease (AD) was only definitively diagnosed after an autopsy. Then, in the 1980s and 1990s, it was discovered that cerebrospinal fluid (CSF) levels of amyloid-β peptide 42 (Aβ42) and tau are correlated with AD neuropathology. In the 2000s, amyloid-binding radiotracers enabled imaging of amyloid plaques by positron emission tomography (PET). More recently, tau radiotracers have revealed the distribution of tangle pathology. Although many investigators worked for decades to develop blood tests for AD, this goal was so elusive that it was referred to as the ‘holy grail’ of AD research. Fortunately, technical improvements in the sensitivity and precision of assays were achieved, and groups began reporting on very promising assays for blood-based AD biomarkers beginning in 2017. Multiple assays have now demonstrated that the ratio of two forms of amyloid-β peptide (Aβ42/Aβ40) in plasma is highly concordant with amyloid PET1,2,3. Plasma levels of phosphorylated tau (P-tau) isoforms, including tau phosphorylated at 181 (P-tau181) and 217 (P-tau217), are not only highly concordant with amyloid PET, but are also associated with the degree of cognitive impairment4,5,6,7,8. Neurofilament light (NfL), which appears to be elevated in many neurological disorders, may also be helpful in predicting cognitive impairment9,10. It seems likely that a single blood sample will soon allow AD to be diagnosed with high confidence in living individuals.
While most studies have evaluated each analyte in isolation, a study in this issue of Nature Aging by Cullen et al. investigated whether combining measures of three blood-based biomarkers better predicts cognitive decline and progression of individuals with mild cognitive impairment (MCI) to AD dementia11. The study examined whether individuals progressed to AD dementia over four years, as predicted by plasma Aβ42/Aβ40, P-tau181 and NfL. Interestingly, the best performing model included plasma P-tau181 and NfL (area under the receiver operating curve of 0.88), but did not improve with addition of Aβ42/Aβ40. This finding makes sense within the context of a large body of work in AD suggesting that biomarkers of tau and neurodegeneration are more tightly linked to cognition than biomarkers of amyloid. However, models with Aβ42/Aβ40 and NfL performed almost as well, suggesting that Aβ42/Aβ40 and P-tau181 provide redundant information, while NfL provides complementary information to both Aβ42/Aβ40 and P-tau181. Notably, the model incorporating P-tau181 and NfL performed similarly to CSF biomarkers and was superior to a model of covariates without blood-based biomarkers. In future studies, it will be important to determine which combination of blood-based measures best predicts progression from cognitive normality, rather than MCI, to very mild AD dementia; it is possible that plasma Aβ42/Aβ40 may improve prediction of progression at this earlier stage of disease. A strength of the Cullen et al. study is that analyses were initially performed in the Swedish BioFINDER study and then validated in the Alzheimer’s Disease Neuroimaging Initiative study. Additionally, Cullen et al. created an online calculator that incorporates blood-based biomarker levels and individual level factors (age, sex, baseline cognitive performance and biomarker status) to predict prognosis. Models such as these may provide more accurate prediction of future AD dementia risk compared to current approaches, which often apply a single, dichotomous biomarker cut-off (for example, for amyloid PET status) to stratify AD dementia risk, regardless of individual characteristics (Fig. 1).
The most immediate use of AD dementia risk models that include blood-based biomarkers is in clinical trials, where biomarkers have already become essential tools. In some AD clinical trials conducted before AD biomarkers were available, approximately one-quarter of participants did not have AD brain pathology but rather had non-AD causes of cognitive impairment12. Now, most AD clinical trials use AD biomarkers to ensure that enrolled participants have AD brain pathology. Further, many clinical trials use biomarkers to monitor the effects of treatments, which allows investigators to confirm that treatments are having the intended biological effects13. AD prevention clinical trials implement interventions during the preclinical phase, when individuals have early AD brain pathology but are still cognitively normal. The preclinical phase may represent a critical period for intervention, when individuals at risk for dementia can be identified by AD biomarkers but have not yet experienced the extensive neurodegeneration that may make modification of the disease course more difficult. AD prevention trials require screening large numbers of cognitively normal individuals for AD brain pathology using AD biomarkers14. Prognostic models for AD dementia risk based on combinations of biomarkers such as the one introduced by Cullen et al. may increase the efficiency of prevention trials by improving identification of individuals who are likely to experience cognitive decline over the study period.
AD clinical trials would greatly benefit from the practical advantages of blood-based biomarkers compared to biomarkers that require CSF collection or advanced imaging techniques. Because they are less invasive, do not require highly skilled personnel or specialized facilities and are likely to be less expensive than CSF biomarkers or amyloid PET, blood tests would be ideal for screening participants for enrollment in AD clinical trials. The potential benefits are multiplied in prevention trials, where the percent of screen failures approach 60–80% due to the lower frequency of AD brain pathology in cognitively normal individuals14. Further, it may be possible for potential participants to undergo blood collection at a lab near their home and have the sample shipped to a central lab for biomarker measurement, which could eliminate the in-person visit for biomarker ascertainment that currently represents a major bottleneck in study enrollment. It is also possible that groups traditionally underrepresented in AD clinical trials, such as African Americans, would be more willing to undergo a blood test than other procedures, which could enable treatments to be tested in a more inclusive cohort. It is currently unknown whether blood-based biomarkers can be used to monitor treatment effects, but given the relatively high correlation of plasma and CSF biomarkers, it seems likely that blood-based biomarkers will be useful. In sum, blood-based biomarkers of AD have the potential to be a game changer for the development of effective AD treatments.
Models incorporating blood-based biomarkers are also likely to revolutionize the clinical diagnosis of AD dementia. Amyloid PET and CSF biomarkers were validated for clinical use as tests for AD brain pathology rather than AD dementia, and the same cut-off is applied to all individuals. In contrast, models that connect biomarker levels to AD dementia may enable a test for AD dementia rather than AD brain pathology. Clinicians ordering biomarker tests are trying to answer the question of whether an individual has cognitive impairment caused by AD, not whether an individual has AD brain pathology. This distinction is important because many individuals with AD brain pathology are cognitively normal; therefore, even if AD pathology is present, it may not be the cause of cognitive impairment in a patient. Models for AD dementia that incorporate biomarker levels and individual level factors may improve the ability of clinicians to understand whether cognitive impairment is related to AD or other potential causes in individual patients.
An AD blood test could become a routine element of the diagnostic work-up for dementia. Currently, when patients present with significant cognitive concerns, the standard of care is to check blood chemistries, blood cell counts, thyroid function tests and vitamin B12 levels, and to perform structural brain imaging. Only a small fraction (likely <5%) of patients presenting for evaluation of cognitive impairment undergo AD biomarker testing, even though these tests may affect diagnosis and management in many patients15. When accurate AD blood tests are available at a reasonable cost to patients, they may become a standard part of the dementia work-up. Such testing could enable an earlier and more accurate diagnosis of AD dementia, which is desirable even if effective disease-modifying therapies are not yet available. Further, patients found to have AD could be referred to clinical trials, significantly increasing the number of eligible and interested clinical trial participants. Once effective therapies are approved, patients with cognitive concerns will require testing for AD brain pathology so that appropriate treatment may be promptly initiated; a blood test will likely be the most efficient approach. If treatments are found that prevent the onset of dementia symptoms in cognitively normal individuals, screening of all older adults for AD brain pathology with blood tests may become the standard of care. While these goals may seem many years into the future, the recent rapid development of a blood-based biomarker for AD demonstrates that progress can occur suddenly. The holy grail of an AD blood test has been found. Now researchers must use this game-changing tool in the quest for effective AD treatments.
Nakamura, A. et al. Nature 554, 249–254 (2018).
Ovod, V. et al. Alzheimers. Dement. 13, 841–849 (2017).
Schindler, S. E. et al. Neurology 93, e1647–e1659 (2019).
Palmqvist, S. et al. JAMA 324, 772–781 (2020).
Mielke, M. M. et al. Alzheimers. Dement. 14, 989–997 (2018).
Barthelemy, N. R., Horie, K., Sato, C. & Bateman, R. J. J. Exp. Med. 217, e20200861 (2020).
Janelidze, S. et al. Nat. Med. 26, 379–386 (2020).
Karikari, T. K. et al. Lancet Neurol. 19, 422–433 (2020).
Preische, O. et al. Nat. Med. 25, 277–283 (2019).
Mattsson, N., Cullen, N. C., Andreasson, U., Zetterberg, H. & Blennow, K. JAMA Neurol. 76, 791–799 (2019).
Cullen, N. C. et al. Nat. Aging https://doi.org/10.1038/s43587-020-00003-5 (2020).
Karran, E. & Hardy, J. N. Engl. J. Med. 370, 377–378 (2014).
Bateman, R. J. & Klunk, W. E. Neurotherapeutics 5, 381–390 (2008).
Sperling, R. A. et al. Sci. Transl. Med. 6, 228fs213 (2014).
Rabinovici, G. D. et al. JAMA 321, 1286–1294 (2019).
S.E.S. is supported by National Institute on Aging (grant no. K23AG053426).
R.J.B. co-founded C2N Diagnostics. Washington University and R.J.B. have equity ownership interest in C2N Diagnostics and receive royalty income based on technology (stable isotope labeling kinetics and blood plasma assay) licensed by Washington University to C2N Diagnostics. R.J.B. receives income from C2N Diagnostics for serving on the Scientific Advisory Board. Washington University, with R.J.B. as co-inventor, have filed patent applications (US Patent application nos. 16/610,428; 62/898,407; 62/962,296; PCT/US2020/012959) on technology related to the processes described in this article. R.J.B. has received honoraria as a speaker/consultant/advisory board member from Amgen, AC Immune, Eisai, F. Hoffman-LaRoche and Janssen, and reimbursement of travel expenses from AC Immune, F. Hoffman-La Roche and Janssen. S.E.S. declares no competing interests.
About this article
Cite this article
Schindler, S.E., Bateman, R.J. Combining blood-based biomarkers to predict risk for Alzheimer’s disease dementia. Nat Aging 1, 26–28 (2021). https://doi.org/10.1038/s43587-020-00008-0