The cerebrospinal fluid biomarker ratio Aβ42/40 identifies amyloid positron emission tomography positivity better than Aβ42 alone in a heterogeneous memory clinic cohort

Alzheimer’s disease (AD) is the most common form of dementia, accounting for 60−80% of all dementia syndromes [1]. The disease is defined neuropathologically by the presence of extracellular amyloid-beta (Aβ) plaques and neurofibrillary tangles of hyperphosphorylated tau protein [2, 3]. The 2018 National Institute on Aging and Alzheimer’s Association “A/T/(N)” research framework comprises several biological biomarkers for the diagnosis of AD, including low levels of cerebrospinal fluid (CSF) Aβ(1–42) (Aβ42) and cortical amyloid positron emission tomography (PET) as markers of Aβ pathology (labeled “A”); elevated CSF phosphorylated tau 181 (pTau181) and cortical tau PET as markers of fibrillar tau (labeled “T”); and CSF total tau (tTau), [¹⁸F]-fluorodeoxyglucose-PET hypometabolism, and atrophy on magnetic resonance imaging as markers of neurodegeneration or neuronal injury (labeled “(N)”) [4]. The presence of different markers determines each patient’s individual disease status, i.e., whether the patient has Alzheimer’s pathological change only (Aβ pathology: A⁺) or AD (Aβ and fibrillar tau pathology: A⁺T⁺) [4]. Currently, there are two methods to identify Aβ pathology in vivo. Firstly, PET measuring Aβ pathology using various Aβ binding tracers, such as [¹⁸F]-flutemetamol and [¹⁸F]-florbetapir, can be used [5]. There is high concordance between different tracers [6], which have been validated against Aβ pathology measured by histopathological samples [7, 8]. Secondly, Aβ pathology can be identified by measuring CSF Aβ42 concentration, which could provide a reliable, cost-effective, and quick alternative method to amyloid PET as a diagnostic tool [9‐11].

The fully automated Elecsys^® β-Amyloid(1–42) CSF, Phospho-Tau (181P) CSF, and Total-Tau CSF immunoassays (Roche Diagnostics International Ltd, Rotkreuz, Switzerland) are intended for the in vitro quantitative determination of Aβ42, pTau181, and tTau, respectively, in CSF [11, 12], and are Conformité Européen approved for clinical use. The ratios of these individual biomarkers, pTau181/Aβ42 and tTau/Aβ42 (corresponding to A/T and A/(N) ratios in the A/T/(N) research framework), have demonstrated high concordance with amyloid positivity as determined by PET imaging, and both ratios have been shown to perform better at identifying Aβ pathology than using Aβ42 alone, when using the highest Youden index as the cut-off [9, 13, 14]. The pTau181/Aβ42 and tTau/Aβ42 ratios have also been shown to predict future clinical progression in patients with mild cognitive impairment (MCI) [9]. In one study, pTau181/Aβ42 and tTau/Aβ42 ratios, determined using Elecsys CSF immunoassays, demonstrated high concordance with [¹⁸F]-flutemetamol and [¹⁸F]-florbetapir amyloid PET data in two cohorts comprising patients with symptoms of cognitive impairment or AD: the Swedish BioFINDER cohort (N = 277; PET positive = 110/277; PET negative = 167/277; area under the curve [AUC] [95% confidence intervals (CI)]: 94.4% [91.5−97.3] for pTau181/Aβ42; 94.0% [91.0−97.0] for tTau/Aβ42, and 86.5% [82.3−90.7] for Aβ42); and the Alzheimer’s Disease Neuroimaging Initiative cohort (N = 646; PET positive = 347/646; PET negative = 299/646; AUC [95% CI]: 96.3% [95.2−98.0] for pTau181/Aβ42, 96.3% [94.8−97.7] for tTau/Aβ42, and 92.1% [90.0−94.3] for Aβ42). Both pTau181/Aβ42 and tTau/Aβ42 ratios outperformed Aβ42 alone [9]. The findings of the above study suggest that pTau181/Aβ42 and tTau/Aβ42 ratios determined using Elecsys CSF immunoassays have the potential to accurately identify Aβ pathology as measured by amyloid PET positivity in homogenous cohorts [9].

While using pTau181/Aβ42 and tTau/Aβ42 ratios may be considered an acceptable approach to stratify patients with AD from healthy individuals in clinical practice, it may be valuable in research settings to evaluate Aβ pathology independently when applying the A/T/(N) framework, or determining the concordance of CSF biomarkers or biomarker ratio measurements to amyloid PET positivity. To date, CSF Aβ42 alone has not achieved as high concordance rates to amyloid PET as pTau181/Aβ42 and tTau/Aβ42 ratios [9]; however, some studies have demonstrated that the Aβ42/40 ratio may be accurate in determining amyloid positivity. In 100 non-demented patients with symptoms of cognitive impairment from the Swedish BioFINDER cohort, Aβ42/40 or Aβ42/38 ratios measured using high-resolution mass spectrometry showed high concordance with [¹⁸F]-flutemetamol PET (AUC [95% CI]: 0.85 [0.78−0.93] for Aβ42; 0.95 [0.90−1.00] for Aβ42/40; and 0.94 [0.88−0.99] for Aβ42/38) [10]. Additionally, another study in 215 non-demented patients with symptoms of cognitive impairment showed that Aβ42/40 and Aβ42/38 ratios, measured using three immunoassays (Euroimmun, Germany], Meso Scale Discovery [Rockville, USA], and Quanterix [Billerica, USA]), predicted abnormal amyloid PET using [¹⁸F]-flutemetamol more accurately than Aβ42 alone; AUC values were 0.912−0.975 for Aβ42/40 and Aβ42/38 ratios, compared with 0.810−0.916 for Aβ42 alone [15]. In a retrospective study of 198 mainly cognitively normal or MCI participants, the Aβ42/40 ratio (determined using CSF assays from different manufacturers) achieved a comparable concordance with amyloid PET, compared with Aβ42 alone or pTau/Aβ42 and tTau/Aβ42 ratios calculated using the highest Youden index as the cut-off [13]. Another retrospective study of 202 predominantly cognitively normal individuals and patients with MCI or AD showed comparable concordance with amyloid PET for the CSF biomarker ratios Aβ42/40, pTau/Aβ42, and tTau/Aβ42; all ratios, calculated using the highest Youden index as the cut-off, were superior to using the individual biomarker Aβ42 alone [14]. Moreover, a retrospective study in a mixed cohort (N = 94) using a different automated platform showed comparable concordance with amyloid PET for the CSF biomarker ratios Aβ42/40, pTau/Aβ42, and tTau/Aβ42; all biomarker ratios, calculated using the highest Youden index as the cut-off, were superior to using the individual biomarker Aβ42 alone [16].

In this study, we evaluated the clinical performance from a research perspective of the Aβ42/40 ratio in detecting amyloid positivity as determined by PET, compared with Aβ42 alone, and the pTau181/Aβ42 and tTau/Aβ42 ratios, calculated using fully automated CSF immunoassays (Roche Diagnostics International Ltd) based on pre-defined (determined from Elecsys CSF immunoassay manufacturer instructions) and Youden-derived cut-offs. A heterogeneous cohort of patients with a range of cognitive disorders, who underwent lumbar puncture and amyloid PET, were selected for inclusion, as we aimed to understand how CSF immunoassays may perform in a broad context of pathologies reflecting the typical patient population of a memory clinic.

Both imaging and CSF biomarkers have been identified as valid diagnostic tools in the most recent A/T/(N) research framework as the focus of diagnostic techniques for AD shifts from confirming the presence of symptomatic AD to identifying patients at an earlier stage of the disease [4]. Previous studies evaluating the performance of Elecsys CSF immunoassays approved for clinical use have shown high concordance for pTau181/Aβ42 and tTau/Aβ42 ratios with amyloid PET in different cohorts and demonstrated that these ratios perform better than the individual biomarkers alone [9, 13, 14].

In this study, we evaluated the performance of fully automated, research-use CSF immunoassays to determine whether they perform as well as the Elecsys CSF assay versions approved for clinical use. All individual CSF biomarkers and biomarker ratios demonstrated good concordance with amyloid PET for the detection of amyloid positivity in a heterogeneous cohort of patients with a broad range of cognitive disorders, extending on previous evidence in more homogeneous cohorts. Our results showed that the CSF Aβ42/40 ratio had higher concordance with amyloid PET than Aβ42 alone and demonstrated comparable performance to the pTau181/Aβ42 and tTau/Aβ42 ratios. The Aβ42/40 ratio also demonstrated robust differentiation between amyloid PET-positive and PET-negative patients, which was comparable to that observed using pTau181/Aβ42 and tTau/Aβ42 ratios.

Consistent with Janelidze et al., our findings suggest that the CSF Aβ42/40 ratio may better differentiate amyloid-positive from amyloid-negative individuals compared with Aβ42 alone [15]. The high concordance of the Aβ42/40 ratio to amyloid PET observed in our study is comparable to previous studies that used mass spectrometry, platforms other than those used here, or cohorts including healthy controls and patients with AD [10, 13‐15].

A possible reason for the CSF Aβ42/40 ratio outperforming Aβ42 in our study, which was attributable primarily to a higher specificity, is the normalization of interpatient variability in Aβ production, as the level of Aβ peptide production differs between individuals. As a result, it is possible to obtain false-positive results if only Aβ42 is assessed. The Aβ42/40 ratio may compensate for this by assessing two Aβ peptides, of which the latter, Aβ40, is less affected by AD [10, 13‐15, 21]. Additionally, the Aβ42/40 ratio is considered to be more robust than Aβ42 alone against various pre-analytical influences that can lead to false-positive results [22].

By using ratios of one protein to any other brain-derived protein, physiological fluctuations might be compensated for [9]. A/(N) ratios may aid in the normalization of patient differences in Aβ production and increase the robustness of CSF biomarkers [23]; however, the level of tTau in CSF represents the severity of neurodegeneration only, which is not specific to AD [23]. tTau can also be elevated in other neurological diseases, such as Creutzfeldt-Jakob disease or following a stroke [24, 25]. Therefore, determining the tTau/Aβ42 ratio may result in an inaccurate diagnosis of AD; for example, determining tTau/Aβ42 ratio in patients with MCI may result in a false-positive diagnosis of AD as a result of elevated tTau, despite normal levels of Aβ. CSF Tau levels are dependent on the stage of the disease and the rate of disease progression, which may lead to errors in CSF biomarker interpretation as these associations are inconsistent and not yet fully understood [26, 27]. Moreover, studies have demonstrated that almost half of AD patients do not exhibit abnormal tau, nor elevated CSF tTau [27].

In agreement with previous studies [9, 10], our findings suggest that CSF biomarker analysis may be a valid alternative to PET for the prediction of amyloid positivity in the diagnosis of a range of cognitive disorders, including AD. In general, using CSF analysis for the diagnosis of AD would reduce the time and expense spent on PET examinations. Patients would not be exposed to PET radioactivity and CSF assessments could take place at non-specialized centers [9], which would increase patient accessibility to accurate diagnosis of AD. Moreover, CSF analysis could allow patients to be easily assessed and categorized in accordance with the A/T/(N) research framework (e.g., using fully automated CSF immunoassays to enable assessment of A/T/(N) from the same patient sample) [4], which would aid the diagnosis of patients with AD, particularly those who may be asymptomatic, and the identification of patients suitable for disease-modifying treatments, such as anti-amyloid immunotherapies. On the other hand, amyloid PET examinations not only determine the presence of amyloid positivity in patients, but also quantify the regional amyloid load in the brain, which may be desirable to evaluate the effect of novel anti-amyloid treatments that are currently in development for AD [28]. Therefore, CSF biomarkers may offer an alternative tool to PET in predicting amyloid positivity, but only for global amyloid quantification in the diagnostic pathway of a range of cognitive disorders, including AD.

We demonstrate that the CSF Aβ42/Aβ40 ratio measured using fully automated, research-use CSF immunoassays can predict amyloid positivity as determined by PET with high accuracy in patients with a broad range of cognitive disorders reflecting a typical heterogeneous population of a memory clinic. The performance of the Aβ42/40 ratio in detecting amyloid positivity by PET was higher than using Aβ42 alone and comparable with the performance of the pTau181/Aβ42 and tTau/Aβ42 ratios. Our results add to previously reported literature demonstrating the strong clinical performance of pTau181/Aβ42 and tTau/Aβ42 ratios in detecting amyloid positivity in homogenous cohorts. Taken together, these findings suggest CSF biomarker analysis could provide a cost-effective alternative to PET in the AD diagnostic pathway.