Concordance between the updated Elecsys cerebrospinal fluid immunoassays and amyloid positron emission tomography for Alzheimer’s disease assessment: findings from the Apollo study

Introduction

Alzheimer’s disease (AD) is a progressive brain disease accounting for 60–80 % of dementia cases in the United States [1], 2]. Globally, the prevalence of AD and other dementias is estimated to increase from 57.4 million cases in 2019, to 152.8 million by 2050 [3].

AD pathology involves accumulating amyloid-β (Aβ) plaques and the hyperphosphorylation of tau proteins (pTau) [4]. The recent development of disease-modifying treatments targeting the pathophysiology of AD, such as donanemab and lecanemab, has highlighted the need for accurate diagnostic tests [5], [6], [7], [8]. Recommendations provided by the International Working Group on the clinical diagnosis of AD suggest that the assessment of biological parameters, such as cerebrospinal fluid (CSF) biomarkers and amyloid positivity by positron emission tomography (PET) imaging, may help detect biological changes before symptom onset and aid in early AD diagnosis [9].

According to recent AD diagnostic criteria, the levels of β-amyloid(1–42) (Aβ₄₂), tau phosphorylated at a threonine residue at position 181 (pTau₁₈₁) and total tau (tTau) in CSF play a crucial role in the timely and accurate diagnosis of AD [10], 11]. Aβ₄₂ levels are inversely correlated with amyloid plaque burden, while pTau₁₈₁ and tTau levels are markers for tangle formation and neuronal degeneration, respectively [4]. Early-stage studies have shown that the ratios of Aβ₄₂ with pTau₁₈₁ and tTau may have increased performance in predicting clinical decline and cognitive impairment in AD, compared with each biomarker alone [12], [13], [14], [15]. Although the levels of β-amyloid(1–40) (Aβ₄₀) have been found to remain unaltered in AD [16], the CSF Aβ₄₂/Aβ₄₀ ratio has also demonstrated better diagnostic performance than Aβ₄₂ alone [16], [17], [18], [19], [20].

The fully automated Elecsys^® β-Amyloid (1–42) CSF, Elecsys Phospho-Tau (181P) CSF and Elecsys Total-Tau CSF immunoassays (Roche Diagnostics International Ltd, Rotkreuz, Switzerland) are in vitro diagnostic (IVD)-certified electrochemiluminescence immunoassays that employ a quantitative sandwich principle and were developed to aid amyloid pathology detection [21]. Since the initial clinical validation, all three immunoassays have been updated resulting in second-generation immunoassays (Elecsys β-Amyloid (1–42) CSF II [Aβ₄₂ Gen2], Elecsys Phospho-Tau (181P) CSF [pTau] and Elecsys Total-Tau CSF [tTau]), which are IVD-certified for their intended use, have higher thresholds for biotin interference and run on a broader range of analyzers than previously [21]. The updated immunoassays have also been recently approved by the US Food and Drug Administration (FDA) due to their concordance with amyloid PET visual read status and ability to identify the presence of amyloid pathology [22], 23].

The initial Elecsys CSF immunoassay clinical cutoff values were established using CSF samples stored at −80 °C in a research setting [21]. To suit clinical routine testing requirements, a new, simplified pre-analytical procedure has been developed to ensure standardization and reduce pre-analytical variability when handling fresh CSF samples [24]. The new handling procedure and updated Elecsys CSF immunoassays, when used in combination, offer improved robustness in measuring CSF biomarkers [21]. However, due to the susceptibility of Aβ₄₂ to differences in pre-analytical handling, a shift in Aβ₄₂ levels is expected when different protocols are applied [25]. Therefore, the clinical cutoff values for the updated Elecsys Aβ₄₂ Gen2 immunoassay and its ratios with pTau, tTau and Aβ₄₀ were adjusted accordingly, as previously published [21].

The present study aimed to support clinical performance verification of the adjusted cutoffs of the Elecsys immunoassays in terms of their ability to correctly identify patients with subjective cognitive decline (SCD) and mild cognitive impairment (MCI) based on amyloid PET results.

Materials and methods

Results

Concordance of pTau/Aβ₄₂, tTau/Aβ₄₂, Aβ₄₂/Aβ₄₀ and Aβ₄₂ in fresh CSF samples with amyloid PET visual read status

Baseline demographics and clinical characteristics

The Apollo study initially included 108 individuals selected from the BioFINDER-2 cohort based on the criteria described in the Supplementary material. Of the 108 individuals, 16 were excluded due to missing CSF biomarker measurement data and one was excluded during the monitoring process due to not fulfilling the inclusion criterion for Mini-Mental State Examination score (≥24) (Figure 1). Thus, data from 91 individuals were included in the primary analysis.

Figure 1:

Enrollment summary. ^aThe low number of frozen samples fulfilling the required conditions was due to many samples being excluded for exceeding 13 weeks of storage during the COVID-19 pandemic. Aβ₄₀, β-amyloid(1–40); Aβ₄₂, β-amyloid(1–42); CSF, cerebrospinal fluid; PET, positron emission tomography; pTau, phosphorylated tau; tTau, total tau.

Individuals were enrolled from the Swedish BioFINDER-2 study at baseline (61/91; 67.0 %) or at the 2-year follow-up visit (30/91; 33.0 %). The demographic and clinical characteristics of the primary analysis population are summarized in Table 1. The primary analysis population comprised 40 (44.0 %) amyloid PET-positive and 51 (56.0 %) amyloid PET-negative individuals according to the majority vote of three independent readers.

Table 1:

Demographic and clinical characteristics of the primary analysis population, total and split by amyloid PET visual read status.

	PET (visual)-positive (n=40)	PET (visual)-negative (n=51)	Total (n=91)
Age, years, mean (min–max)	72.9 (55.0–90.0)	68.2 (43.0–85.0)	70.3 (43.0–90.0)
Education, years, mean (min–max)	13.2 (7.0–31.0)	12.8 (7.0–22.0)	13.0 (7.0–31.0)
MMSE score, mean (min–max)	28.2 (25.0–30.0)	28.7 (24.0–30.0)	28.5 (24.0–30.0)
Sex, n (%)  Female  Male	15 (37.5) 25 (62.5)	23 (45.1) 28 (54.9)	38 (41.8) 53 (58.2)
SCD/MCI, n (%)  SCD  MCI  Missing	20 (50.0) 15 (37.5) 5 (12.5)	11 (21.6) 26 (51.0) 14 (27.5)	31 (34.1) 41 (45.1) 19 (20.9)
APOE genotype, n (%)  E2/E2  E2/E3  E2/E4  E3/E3  E3/E4  E4/E4	1 (2.5) 2 (5.0) 0 (0.0) 9 (22.5) 21 (52.5) 7 (17.5)	1 (2.0) 3 (5.9) 1 (2.0) 29 (56.9) 16 (31.4) 1 (2.0)	2 (2.2) 5 (5.5) 1 (1.1) 38 (41.8) 37 (40.7) 8 (8.8)
Family history, n (%)  Yes  No  Missing	17 (42.5) 20 (50.0) 3 (7.5)	25 (49.0) 23 (45.1) 3 (5.9)	42 (46.2) 43 (47.3) 6 (6.6)
Visit during the BioFINDER-2 study, n (%)  Baseline visit  2-year follow-up	30 (75.0) 10 (25.0)	31 (60.8) 20 (39.2)	61 (67.0) 30 (33.0)

1. APOE, apolipoprotein E; MCI, mild cognitive impairment; MMSE, Mini-Mental State Examination; PET, positron emission tomography; SCD, subjective cognitive decline.

Amyloid PET concordance analysis

Amyloid PET concordance analysis showed that the performance of the pTau/Aβ₄₂ and tTau/Aβ₄₂ ratios at the adjusted cutoffs was as expected and the pre-defined acceptance criteria were met (pTau/Aβ₄₂: PPA 0.800, NPA 0.882; tTau/Aβ₄₂: PPA 0.775, NPA 0.902; Figure 2; Table 2). The observed concordance between Aβ₄₂/Aβ₄₀, dichotomized at the previously published adjusted cutoff, and amyloid PET status was comparable with a PPA of 0.950 and an NPA of 0.824 (Figure 2).

Figure 2:

Joint distributions of the single biomarkers (A) pTau and Aβ₄₂, (B) tTau and Aβ₄₂ and (C) Aβ₄₂ and Aβ₄₀. Red lines indicate the respective cutoffs (pTau/Aβ₄₂>0.023; tTau/Aβ₄₂>0.28; Aβ₄₂/Aβ₄₀<0.052). Points are coloured by amyloid PET visual read status. Aβ₄₀, β-amyloid(1–40); Aβ₄₂, β-amyloid(1–42); PET, positron emission tomography; pTau, phosphorylated tau; tTau, total tau.

Table 2:

Hypothesis testing of the pre-specified acceptance criteria for pTau/Aβ₄₂, tTau/Aβ₄₂ and Aβ₄₂.

Performance measure	Point estimate (95 % CI)	Acceptance criteria	Testing result
pTau/Aβ42  PPA  NPA	0.800 (0.652–0.895) 0.882 (0.766–0.945)	PPA >0.75 & LCL >0.60 PPA >0.75 & LCL >0.60	Successful Successful
tTau/Aβ42  PPA  NPA	0.775 (0.625–0.877) 0.902 (0.790–0.957)	PPA >0.75 & LCL >0.60 PPA >0.75 & LCL >0.60	Successful Successful
Aβ42  PPA  LR–	0.975 (0.871–0.996) 0.039 (0.006–0.270)	PPA >0.75 & LCL >0.60 UCL <1	Successful Successful

1. Aβ₄₂, β-amyloid(1–42); CI, confidence interval; LCL, lower confidence limit; LR–, negative likelihood ratio; NPA, negative percent agreement; PPA, positive percent agreement; pTau, phosphorylated tau; tTau, total tau; UCL, upper confidence limit.

Using the CSF pTau/Aβ₄₂-based classification, 38 individuals were scored as CSF-positive, of whom 32 were concordant with a positive PET result; 53 individuals were scored as CSF-negative, of whom 45 were concordant with a negative PET result. In total, 77/91 (84.6 %) individuals showed concordant CSF and amyloid PET visual read results (Table 3). Of the 14 individuals with discordant results, eight were CSF-negative with a positive PET result and six were CSF-positive with a negative PET result (Supplementary Table 1); for 6/8 and 2/6 individuals, biomarker values were within ±10 % of the cutoff value (0.023), respectively (Supplementary Table 2). Similar results were observed using the CSF tTau/Aβ₄₂-based classification, where in total 77/91 (84.6 %) individuals showed concordant CSF and amyloid PET visual read results, while 9/14 were CSF-negative with a positive PET result and 5/14 were CSF-positive with a negative PET result (Table 3; Supplementary Table 1). For 5/9 and 1/5 individuals with discordant results, biomarker values were within ±10 % of the cutoff value (0.28), respectively (Supplementary Table 2). Using the exploratory Aβ₄₂/Aβ₄₀-based classification, in total 80/91 (87.9 %) individuals showed concordant CSF and amyloid PET visual read results, while 2/11 were CSF-negative with a positive PET result and 9/11 were CSF-positive with a negative PET result (Table 3; Supplementary Table 1). Two of the nine individuals with CSF-positive and PET-negative results had biomarker values within ±10 % of the cutoff value (0.052) (Supplementary Table 2).

Table 3:

Concordance tables of classification based on pTau/Aβ₄₂, tTau/Aβ₄₂, Aβ₄₂/Aβ₄₀ and Aβ₄₂ vs. amyloid PET visual read status.

	PET (visual)-positive (n=40)	PET (visual)-negative (n=51)	Total (n=91)
pTau/Aβ42  CSF-positive, n (%)  CSF-negative, n (%)	32 (35.2) 8 (8.8)	6 (6.6) 45 (49.5)	38 (41.8) 53 (58.2)
tTau/Aβ42  CSF-positive, n (%)  CSF-negative, n (%)	31 (34.1) 9 (9.9)	5 (5.5) 46 (50.5)	36 (39.6) 55 (60.4)
Aβ42/Aβ40  CSF-positive, n (%)  CSF-negative, n (%)	38 (41.8) 2 (2.2)	9 (9.9) 42 (46.2)	47 (51.6) 44 (48.4)
Aβ42  CSF-positive, n (%)  CSF-negative, n (%)	39 (42.9) 1 (1.1)	18 (19.8) 33 (36.3)	57 (62.6) 34 (37.4)

1. Aβ₄₀, β-amyloid(1–40); Aβ₄₂, β-amyloid(1–42); CSF, cerebrospinal fluid; PET, positron emission tomography; pTau, phosphorylated tau; tTau, total tau. The cutoffs for CSF-positivity were as follows: pTau/Aβ₄₂ >0.023; tTau/Aβ₄₂ >0.28; Aβ₄₂/Aβ₄₀ <0.052; Aβ₄₂ ≤1,030 ng/L.

The concordance analysis also showed that the performance of Αβ₄₂ as a single biomarker at the adjusted cutoff was as expected and met the pre-defined acceptance criteria (PPA 0.975, LR– 0.039; Figure 3; Table 2). Of the 91 individuals tested, 57 individuals were scored as CSF-positive, of whom 39 were concordant with a positive PET result; 34 individuals were scored as CSF-negative, of whom 33 were concordant with a negative PET result. In total, 72/91 (79.1 %) individuals had concordant CSF and amyloid PET visual read results using the updated Αβ₄₂ Gen2 immunoassay (Table 3). The NPA observed for the Αβ₄₂ Gen2 immunoassay was lower than the NPA of the pTau/Aβ₄₂ and tTau/Aβ₄₂ ratios, as expected, and 18/91 (35.3 %) individuals with negative PET scans were misclassified as positive by the Αβ₄₂ immunoassay (Table 3; Supplementary Tables 1 and 2). Nevertheless, the performance of pTau/Aβ₄₂, tTau/Aβ₄₂ and Aβ₄₂ met the pre-specified acceptance criteria for all three biomarkers (Table 2).

Figure 3:

Box plot of Aβ₄₂ concentration (ng/L) by visual PET status. Red lines indicate the respective cutoff (≤1,030 ng/L). Aβ₄₂, β-amyloid(1–42); PET, positron emission tomography.

Stability analysis in frozen CSF samples

Of the 91 CSF samples, 33 samples (36.3 %) were frozen at −20 °C for 1–13 weeks and used to explore the effect of storage and one freeze-thaw cycle on the stability of frozen CSF samples; for Aβ₄₀, measurements were available for 28/33 samples. For all four biomarkers, the measurements in samples before and after freezing were highly correlated, with Pearson’s R >0.99, and slope estimates were close to 1.000 (pTau: 0.973; tTau: 0.965; Aβ₄₂: 1.000; Aβ₄₀: 1.050; Figure 4, Supplementary Table 3). The largest bias estimate at the pre-specified concentration was observed for tTau (bias: −2.74 % [95 % CI –3.42; −1.17]) at a concentration of 300 ng/L, followed by Aβ₄₂ (bias: −2.64 % [95 % CI –6.08; −1.11]), Aβ₄₀ (bias: −1.6 % [95 % CI –5.6; 0.4]) and pTau (bias: −1.15 % [95 % CI –3.94; 0.62]) (Figure 4, Supplementary Table 3). The concentration recoveries for pTau and tTau were within 100 ± 10 % in all samples stored for 1–13 weeks (Figure 5). Aβ₄₂ and Aβ₄₀ recoveries were within 100 ± 10 % in all samples stored at −20 °C for 1–8 weeks (n=6). For Aβ₄₂ and Aβ₄₀, concentration recoveries for 6/27 and 2/22 samples, respectively, stored at the same temperature for >8–13 weeks, were below 90 %.

Figure 4:

Stability of (A) pTau, (B) tTau, (C) Aβ₄₂ and (D) Aβ₄₀ at −20 °C for 1–8 and >8–13 weeks. Passing–Bablok regression fit is shown as a black line with 95 % confidence bounds (light blue shaded area). X-axes show concentrations in fresh samples and y-axes concentrations in frozen samples. Red dashed lines represent identity lines. Blue points indicate storage for 1–8 weeks and green points storage for >8–13 weeks. Aβ₄₀, β-amyloid(1–40); Aβ₄₂, β-amyloid(1–42); pTau, phosphorylated tau; tTau, total tau.

Figure 5:

Concentration recoveries (%) for (A) pTau, (B) tTau, (C) Aβ₄₂ and (D) Aβ₄₀ observed after storage at −20 °C for 1–8 weeks and >8–13 weeks. Red dashed lines indicate 90 and 110 % recovery bounds. Aβ₄₀, β-amyloid(1–40); Aβ₄₂, β-amyloid(1–42); pTau, phosphorylated tau; tTau, total tau.

Discussion

This study supports the concordance of pTau/Aβ₄₂, tTau/Aβ₄₂ and Aβ₄₂ with amyloid PET visual reads and verifies that the performance of the adjusted cutoffs for the Elecsys ratios is as expected in CSF samples handled with the new routine-use pre-analytical procedure and measured with the updated CSF immunoassays. These results suggest that both biomarker ratios plus Aβ₄₂ alone could be used in clinical practice as reliable alternatives to amyloid PET imaging to aid in the diagnosis of amyloid pathology.

In this study, the pTau/Aβ₄₂ and tTau/Aβ₄₂ ratios met the pre-defined acceptance criteria for PPA and NPA and showed more than 80 % concordant positive and negative CSF and PET scan results, while only a low percentage (<16 %) were discordant. Aβ₄₂ alone also met the pre-defined acceptance criteria, and the LR– value was low (0.039). This indicated that the likelihood of an amyloid PET-positive individual having an Aβ₄₂ concentration greater than 1,030 ng/L is significantly smaller (by a factor of 0.039) compared with an amyloid PET-negative individual. Nevertheless, the NPA value of Αβ₄₂ as a single biomarker was substantially lower than the ratios, as biomarkers alone typically perform worse than combination ratios, and the original Αβ₄₂ cutoff value was set to fulfill a high PPA to ensure high sensitivity [29], [30], [31]. The exploratory analysis also indicated that normalization with Aβ₄₀ improved the performance of Aβ₄₂ alone, and the performance of the Aβ₄₂/Aβ₄₀ ratio was comparable to that of the pTau/Aβ₄₂ and tTau/Aβ₄₂ ratios, consistent with previously reported results [32], [33], [34], [35], since the CIs of PPA and NPA overlapped.

Storage at −20 °C for 1–8 weeks and one freeze-thaw cycle had no effect on any of the four biomarker concentration recoveries. Storage for >8–13 weeks also had no significant effect on pTau and tTau concentration recoveries, whereas a small effect was observed on Aβ₄₂ and Aβ_40, respectively, under the same conditions, with 6/27 and 2/22 samples showing concentration recoveries <90 %. It is therefore recommended to store CSF samples at −20 °C for ≤8 weeks to maintain stability.

This study supports the verification of the clinical performance of the updated Elecsys CSF immunoassays with the new routine-use pre-analytical procedure and their concordance with amyloid PET visual read status in distinguishing amyloid-positive individuals with early-stage AD, who are considered perhaps the most relevant but also the most diagnostically challenging group of the intended use population. Although early-stage disease PET scans are challenging to correctly classify as positive or negative, and biomarker levels are closer to the cutoff values, this study indicated a good performance of concordance with amyloid PET imaging. A better performance of the tests in terms of PPA and NPA is expected in individuals with Alzheimer’s dementia (not included here) due to the more advanced amyloid pathology, which is more clearly reflected in the CSF biomarker levels and PET scans.

This study’s findings are consistent with previous research using Elecsys and other platforms. Specifically, previous studies have shown that the Elecsys CSF pTau/Aβ₄₂, tTau/Aβ₄₂ and Aβ₄₂/Aβ₄₀ ratios, as well as Aβ₄₂ alone, are strongly concordant with PET imaging assessing Aβ burden in AD, supporting the use of CSF biomarkers in early amyloid identification [15], 34]. For instance, Hansson et al. indicated that pTau/Aβ₄₂ and tTau/Aβ₄₂, measured with the first-generation Elecsys CSF immunoassays, were highly concordant with amyloid PET visual reads across two different cohorts (BioFINDER and ADNI) comprising different populations and PET radiotracers [15]. Schindler et al. reported high concordance between Pittsburgh compound B PET imaging and pTau/Aβ₄₂, tTau/Aβ₄₂ and Aβ₄₂/Aβ₄₀ ratios, measured using the first-generation Elecsys CSF immunoassays, in discriminating PET-positive from PET-negative individuals [35]. Campbell et al. showed agreement between pTau/Aβ₄₂ and Aβ₄₂/Aβ₄₀ ratios, measured with the first-generation Elecsys and LUMIPULSE immunoassays, and amyloid PET classification, and the biomarker ratio results were superior to individual biomarkers [36]. In another study, Alcolea et al. reported that pTau/Aβ₄₂, tTau/Aβ₄₂ and Aβ₄₂/Aβ₄₀, measured on the fully automated, Conformité Européenne-marked and FDA-approved LUMIPULSE G600II platform (Fujirebio), had good diagnostic agreement with ¹⁸F-flutemetamol amyloid PET and the ratios were suggested to be more reliable in clinical practice than Aβ₄₂ alone [31], 32], 37].

The future clinical application of these findings is expected to aid earlier diagnosis of patients with AD, giving them and their caregivers time to plan for the future and access potential treatments for early symptom management. Implementing the new pre-analytical procedure and recommended storage conditions for handling fresh CSF samples is expected to reduce the variability of assay measurements and enable comparison of CSF biomarker levels between different laboratories, thus increasing the utility of CSF biomarkers in research and routine clinical practice [24].

This study had some limitations, such as the relatively small number of individuals enrolled, which suggests that the results should be confirmed in a wider population. The enrolled population was not randomly selected from the intended use population, but was based on the Swedish BioFINDER-2 study cohort. Thus, the results of this study may be biased due to the inclusion and exclusion criteria of the BioFINDER-2 study. However, the BioFINDER-2 study includes participants from secondary care specialized memory clinics, and therefore does not differ substantially from an intended use population. Additionally, the concordance of Aβ₄₂/Aβ₄₀ with amyloid PET visual reads was assessed using an early version of the Elecsys CSF Aβ₄₀ immunoassay and the acceptance criteria as well as the adjusted cutoff for the Aβ₄₂/Aβ₄₀ ratio were not pre-specified. Moreover, the objectives for the frozen sample analysis were limited to exploratory due to the low number of frozen samples available. The number of samples stored up to 8 weeks was small, suggesting that the results of the exploratory analysis under these storage conditions will need to be confirmed in a larger sample size. It is also worth noting that although the pTau/Aβ₄₂, tTau/Aβ₄₂, Aβ₄₂/Aβ₄₀ ratios and Aβ₄₂ alone can successfully identify individuals with positive amyloid PET results, their performance does not establish a diagnosis of AD or other cognitive disorder and cannot be used for predicting the development of dementia or other neurological conditions, or to monitor responses to therapies.

Conclusions

CSF biomarker status, determined by the pTau/Aβ₄₂, tTau/Aβ₄₂ and Aβ₄₂/Aβ₄₀ ratios and Aβ₄₂ alone in fresh CSF, is concordant with amyloid PET visual read status. All three ratios can be used to identify amyloid PET positivity in individuals with SCD/MCI with high sensitivity and specificity, and Αβ₄₂ alone can distinguish amyloid PET-positive individuals with high sensitivity. As a conservative approach, CSF samples should be stored at −20 °C for ≤8 weeks to maintain stability before testing. The new routine-use pre-analytical procedure and the updated Elecsys Aβ₄₂ Gen2, pTau and tTau CSF immunoassays could be used in clinical practice as alternatives to amyloid PET imaging to identify amyloid positivity in SCD/MCI individuals, thus contributing to the accurate and timely diagnosis of AD.