Statistics from Altmetric.com
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.
Most people would want to know early, if they had Alzheimer’s disease (AD).1 Potential new treatments, such as lecanemab, require early molecular AD diagnosis.2 Yet, very few clinic patients receive a molecular diagnosis (only ~1% in the UK), around 40% of people with dementia are never diagnosed and most diagnoses occur relatively late in the course of the illness.3 Amyloid and tau radionucleotide ligands and cerebrospinal fluid (CSF) AD biomarkers are approved by the USA Food and Drug Administration, adopted in National Institute for Health and Care Excellence (NICE) dementia guidelines, and alter clinical management,4 but they are considered expensive and intrusive, which limits widespread use. Blood biomarkers are a potential solution as they are much easier and cheaper to deliver. In research cohorts, blood biomarkers offer high sensitivity and specificity for AD. However, we require real-world blood biomarker data before validating them for clinical use.5 6
To this end, here, we present three papers that illuminate the practical application of blood biomarkers—Altomare et al 7 (Geneva Cohort) and Lehman et al 8 (BALTAZAR cohort) assess the performance of AD blood biomarkers in memory clinics, and Veria et al 9 highlight potential for positive AD blood biomarkers in amyotrophic sclerosis.
Altomare et al report on 200 patients with cognitive complaints, who range from cognitively unimpaired to dementia stage, to explore the relationship between plasma biomarkers and the established biomarkers of AD including Amyloid (Aß42 or Aß42:40 ratio in the CSF or Amyloid Positron Emission Tomography (PET)), Tau (p-tau181 or p-tau231 in the CSF, or Tau-PET imaging) and markers of Neurodegeneration (atrophy on MRI or hypometabolism on Fluorodeoxyglucose (FDG)-PET).10 The authors rate diagnostic accuracy defined by area under the receiver operated curve (AUC) of 0.7–0.8 as ‘acceptable’ and 0.8–0.9 as ‘excellent’. AUC for some amyloid and tau measures fell within this range (eg, the best amyloid classification AUC was plasma p-tau231:Aß42 compared with CSFAß42=0.87). However, plasma Aß42 compared with amyloid PET had an AUC of 0.66, and, likewise, plasma NFL fell below the acceptable range of diagnostic accuracy (when compared with an MRI score of hippocampal atrophy or FDG-PET).
In the BALTAZAR prospective longitudinal study, 476 participants with mild cognitive impairment (MCI) had baseline plasma p-tau181 and were followed over years for conversion to dementia— arguably the gold-standard outcome diagnostically (as well as a subgroup having baseline CSF Aß assays). Over two-thirds of patients had Amyloidß42:40 ratio<0.1 (Aß+) at baseline and 30% converted to dementia during a period of up to 3 years. Plasma p-tau181 at baseline was associated (1) with baseline CSF Aß+ status (adding P-tau181 to a model containing age, sex, APOEε4 status and Mini-mental State Examination (MMSE) improved AUC from 0.786 to 0.849 (from acceptable to excellent as defined above)) and (2) with conversion to dementia (when added to age, sex, APOEe4 status and MMSE, P-tau181 improved AUC from 0.69 to 0.74 (ie, just into the acceptable range)). MCI participants with low serum p-tau181 (<2.32 pg/mL) showed a conversion rate of <20% over 3 years. Serum creatinine and eGFR were associated with plasma levels of p-tau181—and the authors highlight the importance of correcting for renal function to optimise interpretation of p-tau181.
In our third paper, in a cohort of 148 amyotrophic lateral sclerosis (ALS), 12 spinal muscular atrophy, 88 AD and 70 healthy controls, Veria et al demonstrate raised blood, but not CSF, p-tau181 in ALS—although less marked than in AD. Given the frequency of behavioural and cognitive changes associated with ALS, and the high prevalence of AD, this study demonstrates the impact of comorbidities on biomarkers for neurodegenerative diseases—and that our understanding is still evolving.
Interpretation across these studies has limitations—Altomare’s manuscript has great strengths due to the in-depth characterisation with multiple biomarkers, but heterogeneity in disease stage of the patients, limited reporting of comorbidities, multiple statistical comparisons that are uncorrected and a relatively small sample all limit inference. Lehman et al have a much larger cohort with documented comorbidities, but did not report ptau217,5 an isoform of p-tau thought better for predicting conversion to dementia than ptau181, and patient characterisation was not based on the Amyloid/Tau/Neurodegeneration (ATN) classification11 (eg, there was no biomarker of neurodegeneration reported). Veria et al included a wider range of neurodegenerative diseases but some groups are small (although numbers are impressive given the rarity of the conditions), and comorbidities are little explored. The cognitive tests differ between the studies—Veria not using and/or reporting standardised cognitive tests, while Altomere and Lehman rely on MMSE (a standard cognitive screening tool relatively insensitive to early disease). Also, the ethnicity of participants was not mentioned in the studies.
These studies are important as they show that blood biomarkers in real-world cohorts mirror more established biomarker results and can predict future cognitive decline. But questions remain—which biomarkers should be used and at what stage of disease, how should comorbidities (including renal function) be considered when interpreting blood biomarker results? More fundamentally, do we think that diagnostic accuracy for AD of 0.7–0.8 compared with gold standard tests is acceptable—or is this really an accuracy level we would accept for patient triage only? In other words, will we always require diagnostic confirmation with amyloid PET or CSF before embarking on Alzheimer-specific treatments—or is it reasonable to aim for diagnosis based solely on blood biomarker tests?
The recently published BIODEGMAR cohort12 suggested perhaps blood tests could be diagnostic in themselves—comparison between CSF Aβ42/p-tau and plasma ptau181 achieved AUC=0.94, and p-tau217 achieved AUC=0.94. But our papers, together, suggest we should be cautious. Why do blood biomarkers appear less accurate here? Of course, for the BALTAZAR cohort the main outcome is delayed diagnostic verification, but even the cross-sectional comparison suggests lower discrimination by ptau181 than BIODEGMAR. It is likely that demographic and perhaps technical (eg, assay) might have contributed to differences. Patient selection might also be very important for interpretation of these difference cohorts—comorbidities affect diagnostic accuracy and typical research cohorts have fewer comorbidities than patients in real-world clinics.
We need aligned populations, index and reference tests across studies to generate larger real-world datasets, with in-depth patient characterisation, before we can be confident clinically how to use blood biomarkers. We agree with current clinical recommendations suggesting their cautious use only in specialised memory services and confirmation with CSF/PET.13 Blood biomarkers will transform AD diagnosis, but it looks unlikely they will fully replace more established AD diagnostic biomarkers in the next few years (at least). If memory clinics aspire to deliver disease-modifying therapies in near future, CSF and/or PET scanning will need to become routine.
Patient consent for publication
Contributors AAH and EC contributed equally to this manuscript.
Funding EC: funding from MRC, NIHR, Rosetrees, BRACE, Above and Beyond, Wellcome, Biogen. Acted as a consultant and/or provided educational resources for Biogen, Eisai and Roche. AAH: MD, PhD, funding from the MRC, UK (grant MR/T005580/1); National Institute of Health/National Institute of Aging, USA (grant 1R56AG074467-01); Biogen for teaching.
Competing interests None declared.
Provenance and peer review Commissioned; internally peer reviewed.