RESEARCH ARTICLE Agreement of amyloid PET and CSF biomarkers for Alzheimer’s disease on Lumipulse Daniel Alcolea1,2 , Jordi Pegueroles1,2, Laia Mun ~ oz1,2, Valle Camacho3, Diego Lo  pez-Mora3, Alejandro Fernandez-Leo n , Nathalie Le Bastard , Els Huyck , Alicia Nadal , Vero 3 4 4 5  nica Olmedo5, 2 1,2 Frederic Sampedro , Victor Montal , Eduard Vilaplana 1,2  n , Rafael Blesa1,2, Juan , Jordi Clarimo 1,2 1,2 Fortea & Alberto Lleo  1,2 1 Sant Pau Memory Unit, Department of Neurology, Hospital de la Santa Creu i Sant Pau - Biomedical Research Institute Sant Pau (IIB Sant Pau), Universitat Auto noma de Barcelona, Barcelona, Spain 2 Centro de Investigacio n Biomedica en Red de Enfermedades Neurodegenerativas, CIBERNED, Madrid, Spain 3 noma de Barcelona, Nuclear Medicine Department, Hospital de la Santa Creu i Sant Pau, Biomedical Research Institute Sant Pau,, Universitat Auto Barcelona, Spain 4 Fujirebio Europe N.V., Gent, Belgium 5 Fujirebio Iberia, S.L., Barcelona, Spain Correspondence Abstract Daniel Alcolea, Sant Pau Memory Unit, Department of Neurology, Hospital de la Objective: To determine the cutoffs that optimized the agreement between 18 Santa Creu i Sant Pau. Sant Antoni Maria F-Florbetapir positron emission tomography (PET) and Ab1-42, Ab1-40, Claret 167, 08025 Barcelona, Spain. Tel: tTau, pTau and their ratios measured in cerebrospinal fluid (CSF) on the +34935565986; Fax: +34935568604; E-mail: LUMIPULSE G600II instrument, we quantified the levels of these four dalcolea@santpau.cat biomarkers in 94 CSF samples from participants of the Sant Pau Initiative on Neurodegeneration (SPIN cohort) using the Lumipulse G System with available 18 Funding Information F-Florbetapir imaging. Methods: Participants had mild cognitive impairment This study was supported by CIBERNED and (n = 35), AD dementia (n = 12), other dementias or neurodegenerative diseases Instituto de Salud Carlos III PI13/01532 and (n = 41), or were cognitively normal controls (n = 6). Levels of Ab1-42 were PI16/01825 to R.B.; PI14/01561 and PI17/ standardized to certified reference material. Amyloid scans were assessed visually 01895 to A.L; PI14/01126 and PI17/01019 to and through automated quantification. We determined the cutoffs of CSF J.F., PI18/00435 to D.A., jointly funded by biomarkers that optimized their agreement with 18F-Florbetapir PET and evalu- Fondo Europeo de Desarrollo Regional n Europea, “Una manera de ated concordance between markers of the amyloid category. Results: Ab1-42, (FEDER), Unio hacer Europa”. This work was also supported tTau and pTau (but not Ab1-40) and the ratios with Ab1-42 had good diagnos- by Generalitat de Catalunya (2017-SGR-547, tic agreement with 18F-Florbetapir PET. As a marker of amyloid pathology, the SLT006/17/125 to D.A., SLT006/17/95 to E.V. Ab1-42/Ab1-40 ratio had higher agreement and better correlation with amyloid and SLT006/17/119 to J.F.) and “Marato  PET than Ab1-42 alone. Interpretation: CSF biomarkers measured with the TV3” foundation grants 20141210 to J.F., Lumipulse G System show good agreement with amyloid imaging in a clinical 20161431 to R.B. and 20142610 to A.L. setting with heterogeneous presentations of neurological disorders. Combination All kits and reagents to perform of Ab1-42 with Ab1-40 increases the agreement between markers of amyloid cerebrospinal fluid analyses in this study were provided by Fujirebio Europe. la Caixa pathology. Foundation. Received: 10 May 2019; Revised: 26 July 2019; Accepted: 2 August 2019 Annals of Clinical and Translational Neurology 2019; 6(9): 1815–1824 doi: 10.1002/acn3.50873 ª 2019 The Authors. Annals of Clinical and Translational Neurology published by Wiley Periodicals, Inc on behalf of American Neurological Association. 1815 This is an open access article under the terms of the Creative Commons Attribution License, which permits use, distribution and reproduction in any medium, provided the original work is properly cited. Amyloid PET and CSF Biomarkers in Lumipulse D. Alcolea et al. AD CSF biomarkers using this platform. Our aims were Introduction to determine for the first time the cutoffs that optimized Advances in the field of biomarkers have pushed forward the agreement between 18F-Florbetapir PET and Ab1-42, a redefinition of Alzheimer’s disease (AD) as a biological Ab1-40, tTau, pTau and their ratios measured in CSF on construct.1 Under this definition, the ATN classification the LUMIPULSE G600II instrument, and to evaluate the system recognizes three general groups of biomarkers for concordance between markers of the amyloid category. AD: biomarkers of b-amyloid plaques (A), biomarkers of fibrillar tau (T) and biomarkers of neurodegeneration or Methods neuronal injury (N).1 These categories can be assessed by using different modalities, but those that are more widely Study participants implemented are cerebrospinal fluid (CSF) biomarkers and imaging techniques. Amyloid positron emission We included 94 participants from the Sant Pau Initiative tomography (PET), CSF Ab1-42 and the ratio Ab1-42/ on Neurodegeneration (SPIN cohort) recruited between Ab1-40 correspond to the “A” category, Tau PET and CSF November 2013 and September 2017 who had available pTau to the “T” category, whereas 18F-Fluorodeoxyglu- CSF samples and 18F-Florbetapir PET imaging (Table 1). cose-PET, anatomical MRI and tTau are considered mark- The SPIN cohort is a multimodal research cohort for bio- ers of the “N” category.1 Within the “A” category, the marker discovery and validation that includes participants concordance between CSF Ab1-42 and amyloid PET imag- with different neurodegenerative dementias, mild cogni- ing is high, but not perfect.2–6 In fact, the Ab1-42/Ab1-40 tive impairment and cognitively normal controls. All par- ratio has shown to have better agreement with amyloid ticipants receive an extensive neurological and PET imaging compared to levels of Ab1-42 alone.4,7–9 neuropsychological evaluation and undergo structural 3T In recent years, fully automated platforms have been brain MRI, blood extraction, and lumbar puncture for developed for the analysis of CSF biomarkers. Recently, CSF biomarkers. A subset of participants also receives four CSF analytes (Ab1-42, Ab1-40, tTau and pTau) have molecular imaging such as 18F-Fluorodeoxyglucose-PET, been implemented on the fully automated Lumipulse G amyloid and/or Tau PET. More information on the SPIN System, but there are no validated cutoffs for these four cohort can be found at https://santpaumemoryunit.com/ Table 1. Clinical and demographic characteristics of all participants and based on visual amyloid PET status. All participants Amyloid positive Amyloid negative P value n (%) 94 (100%) 59 (63%) 35 (37%) – Age, years 73.0 (7.6) 73.5 (7) 72.1 (8.5) 0.405* Sex, female/male (% female) 50/44 (53%) 32/27 (54%) 18/17 (51%) 0.960† APOEe4 +/ (% +) 56/37 (60%) 30/28 (52%) 7/28 (20%) 0.004† MMSE score 24.7 (4.1) 24.1 (4.3) 25.7 (3.6) 0.052* Time difference between amyloid PET and lumbar puncture, days 152 (86) 151 (78) 156 (100) 0.792* Clinical diagnosis, n (%) 0.067† Cognitively normal 6 (100%) 3 (50%) 3 (50%) – Mild cognitive impairment 35 (100%) 23 (66%) 12 (34%) – AD dementia 12 (100%) 11 (92%) 1 (8%) – Dementia with Lewy bodies 30 (100%) 18 (60%) 12 (40%) – Frontotemporal dementia 9 (100%) 3 (33%) 6 (67%) – Other diagnoses 2 (100%) 1 (50%) 1 (50%) – CSF biomarkers Ab1-42, pg/mL 745 (379) 608 (213) 985 (474) <0.001* Ab1-40, pg/mL 12252 (3944) 12782 (3954) 11360 (3816) 0.089* tTau, pg/mL 566 (363) 667 (375) 397 (273) <0.001* pTau, pg/mL 93 (71) 114 (71) 58 (58) 0.001* Ab1-42/Ab1-40 0.064 (0.028) 0.049 (0.015) 0.088 (0.027) <0.001* tTau/Ab1-42 0.95 (0.76) 1.17 (0.63) 0.57 (0.81) <0.001* pTau/Ab1-42 0.16 (0.15) 0.20 (0.12) 0.09 (0.16) <0.001* Unless otherwise specified, results are presented as mean (standard deviation). MMSE, Mini-mental state examination; AD, Alzheimer’s disease; CSF, cerebrospinal fluid; PET, positron emission tomography. P-values were calculated by comparing amyloid–positive and amyloid–negative participants using Welch two-sample t-test (*) or Fisher’s exact test (†). 1816 ª 2019 The Authors. Annals of Clinical and Translational Neurology published by Wiley Periodicals, Inc on behalf of American Neurological Association. D. Alcolea et al. Amyloid PET and CSF Biomarkers in Lumipulse our-research/spin-cohort (Alcolea et al., submitted). All The personnel involved in the CSF analyses for this participants gave written consent, and the ethics commit- study were blinded to the clinical diagnosis and to previ- tee of Hospital Sant Pau approved all procedures included ous biomarker determinations. in this study. Amyloid-PET imaging acquisition, visual CSF samples acquisition and analysis assessment and quantitative analysis CSF samples were collected in 10 mL polypropylene tubes All participants underwent amyloid PET imaging with 18 (Sarstedt, Ref#62.610.018) and transferred to the Sant Pau F-Florbetapir as described elsewhere.13 PET data were Memory Unit’s laboratory where they were processed acquired using a Philips Gemini TF scan 50 min after within the first 2 h after acquisition. After centrifugation injection of 370mBq of 18F-Florbetapir. After obtaining (2000g 9 10 min, 4°C), volumes of 0.5 mL of CSF were the transmission data, brain PET dynamic acquisition was aliquoted into polypropylene tubes (Sarstedt, performed (2 9 5 min frames). The reconstruction Ref#72.694.007) and stored at 80°C until analysis. method was iterative (LOR RAMBLA, three iterations and On the day of the analysis, samples were thawed at 33 subsets) with a 128 9 128 image size, 2 mm pixel size room temperature and the tubes were vortexed for 5– and slice thickness. 10 sec. To avoid the effect of multiple freeze-thaw cycles, Three expert readers (V.C., D.L-M. and A.F-L.) blind to aliquots used in this study had not been thawed before. clinical diagnosis and to CSF biomarkers visually rated all Ab1-42, Ab1-40, tTau and pTau were quantified directly PET scans. Following manufacturer’s protocol, scans were from the storage tubes containing 0.5 mL of CSF using classified as “positive” when one or more areas showed the Lumipulse G b-Amyloid 1-42, b-Amyloid 1-40, Total increased cortical gray matter signal resulting in reduced or Tau and pTau 181 assays on LUMIPULSE G600II auto- absent contrast between gray matter and white matter. Scans mated platform (Fujirebio) and following the manufac- were classified as “negative” when the contrast between gray turer’s instructions. We used an adapter to fit the tubes matter and white matter was clear. Final classification as in the equipment. We used the same batch of reagents for “positive” or “negative” was decided upon agreement of at each biomarker throughout the study, and for each sam- least two of three readers. Mean inter-reader overall agree- ple, we measured all four analytes from the same aliquot ment was 88.4% (Min = 87.0%, Max = 90.2%). and in the same run. The platform was configured to We also quantified amyloid deposition. Each partici- start the analysis with Ab1-42, followed by Ab1-40, tTau pant’s PET scan was spatially normalized to a MNI152 18F- and pTau. Buffer-based quality control testing was per- Florbetapir template using a linear and nonlinear transfor- formed at the beginning of each test day to ensure that mation.16 Mean 18F-Florbetapir uptake was measured all measured values of each control level (low, medium across frontal, lateral parietal, lateral temporal and ante- and high) were within the target ranges. rior/posterior cingulate. Then, the 18F-Florbetapir stan- The results of the Lumipulse G b-Amyloid 1-42 pre- dardized uptake value ratio (SUVR) map was extracted sented in this study have been standardized according to using the whole cerebellum as reference.17 The PET scans certified reference material developed by the International of five participants were not suitable for 18F-Florbetapir Federation of Clinical Chemistry and Laboratory Medi- quantification and excluded of quantitative analyses. cine as recommended by their working group for CSF proteins.10 Briefly, values of the calibration standards of Statistical analysis the LUMIPULSE G600II were adapted to the certified ref- erence material resulting in an adjustment of concentra- We performed receiver operating characteristic (ROC) tions that was linearly proportional throughout the range. analysis for Ab1-42, Ab1-40, tTau, pTau and the ratios The aim of standardization to certified reference material Ab1-42/Ab1-40, tTau/Ab1-42 and pTau/Ab1-42 to calcu- is to harmonize immunoassays of Ab1-42 to make results late areas under the curve (AUC) with 95% confidence comparable across different platforms. intervals (DeLong). We compared ROC curves by two- Ab1-42, Ab1-40, tTau and pTau levels in CSF from sided bootstrapping with 2000 replications. For biomark- participants of this study had been measured previously ers and ratios that showed AUC higher than 0.70, we using other immunoassays (INNOTEST b-AMYLOID(1- determined positive percent agreement (PPA or sensitiv- 42), INNOTEST hTAU Ag, and INNOTEST PHOSPHO- ity) and negative percent agreement (NPA or specificity) TAU(181P), Fujirebio Europe; and High Sensitivity Human and calculated optimal cutoffs maximizing their Youden J Amyloid b40, Merck-Millipore), and these results were index (PPA + NPA 1). We calculated overall percent available in our database for their comparison with the agreement (OPA) between CSF biomarker cutoffs and LUMIPULSE analyses.11–15 amyloid PET visual interpretation as the sum of ª 2019 The Authors. Annals of Clinical and Translational Neurology published by Wiley Periodicals, Inc on behalf of American Neurological Association. 1817 Amyloid PET and CSF Biomarkers in Lumipulse D. Alcolea et al. participants classified as “positive” or as “negative” by system, tTau and pTau were highly correlated (Pearson’s both modalities over the total number of participants. We r of 0.98, P < 0.001). also analyzed the correlation of CSF biomarkers with glo- bal amyloid accumulation by fitting quadratic models and Agreement between 18F-Florbetapir visual calculated the agreement of CSF biomarkers cutoffs with status and CSF biomarkers the PET scans quantification status applying a previously described SUVR cutoff of 1.11.17 Level of significance was As displayed in Figure 1A, of individual biomarkers, tTau set at a = 0.05. We used Analyse-it statistical software and pTau had the highest accuracy and showed AUC of 0.80 for the selection of optimal cutoffs and packages “car” (95%CI 0.70–0.89, P < 0.001) and 0.84 (95%CI 0.75–0.93, (v.3.0-3, Fox&Weisberg 2019), “pROC” (v.1.15.0, Robin P < 0.001), respectively. Ab1-42 had fair accuracy with an et al 2011), “grid” (R core team 2019) and “ggplot2” AUC of 0.76 (95%CI 0.65–0.86, P < 0.001) and Ab1-40 (v.3.1.1, Wickham 2016) as implemented in R statistical alone was not useful for the detection of the visual status of software (v 3.6.0) for plots and statistical analyses. amyloid scans (AUC 0.59; 95%CI 0.47–0.71, P = 0.134). Figure 1B shows that the combination of Ab1-42 with a second analyte resulted in significant increases of accu- Results racy. The AUC of Ab1-42/Ab1-40 was 0.86 (95%CI 0.77– 0.96, P < 0.001), higher than that of Ab1-42 alone Study participants (D = 2.5; P = 0.01) or Ab1-40 alone (D = 4.0; We included 94 participants in the study. Table 1 sum- P < 0.001). tTau/Ab1-42 had an AUC of 0.87 (95%CI marizes demographic characteristics and biomarker 0.78–0.95, P < 0.001), higher than that of tTau results. There were no differences in age or sex between (D = 2.2; P = 0.03), and pTau/Ab1-42 had an AUC of both groups. As expected, the amyloid–positive group 0.88 (95%CI 0.79–0.97, P < 0.001), higher than that of had a higher proportion of APOEe4 carriers compared to pTau alone (D = 1.9; P = 0.05). There were no signifi- the amyloid–negative group (52% and 20%, respectively; cant differences in the AUC of the three ratios, and com- P = 0.004). bining a third biomarker in the ratios did not improve their accuracy (data not shown). Quantification of Ab1-42, Ab1-40, tTau and pTau concentrations on the LUMIPULSE CSF biomarker cutoffs based on visual G600II interpretation of amyloid status We measured Ab1-42, Ab1-40, tTau and pTau levels For those biomarkers and ratios that showed AUC higher simultaneously on the Lumipulse G System. Their levels in than 0.70, we used ROC analysis to obtain PPA, NPA and the overall study population ranged from 315 to 2280 pg/ OPA for all possible cutoffs. As displayed in Figure 2, in the mL for Ab1-42, 4585 to 25925 pg/mL for Ab1-40, 141 to case of single biomarkers, Ab1-42, tTau and pTau, the selec- 1902 pg/mL for tTau, and 18 to 340 pg/mL for pTau. tion was based on clear Youden peaks at 916, 456 and The analyses were divided over three calibration runs 63 pg/mL, respectively. For the ratios Ab1-42/Ab1-40, on the LUMIPULSE G600II, and the calibration status tTau/Ab1-42 and pTau/Ab1-42, plots showed plateau stages was valid for all samples. Mean interassay coefficients of indicating that a wide range of cutoffs yielded similar You- variation are displayed in Figure S1. den indices. Best cutoffs for ratios were 0.062 for Ab1-42/ Most CSF samples included in this study had previ- Ab1-40, 0.62 for tTau/Ab1-42, and 0.068 for pTau/Ab1-42. ously been analyzed using other immunoassays, and their Figure 3 displays the agreement between visual status of 18 results were available in our database. Although these his- F-Florbetapir and CSF biomarker cutoffs. For Ab1-42, toric results were obtained in the context of routine clini- tTau and pTau, the OPA values between visual status and cal assessment by using different batches, and therefore a CSF biomarkers status were 79%, 78% and 81%, respec- side-to-side precision analysis could not be performed, we tively. The ratio of Ab1-42 with Ab1-40, tTau and pTau explored their correlation with the Lumipulse G quantifi- increased the OPA to 84%, 82% and 88%, respectively. cations. The Lumipulse G assays for Ab1-42, tTau and pTau showed very high correlation with values previously Markers of amyloid and importance of measured with Fujirebio’s INNOTEST ELISA (Pearson’s r assessing a second biomarker of 0.94, 0.95 and 0.95, respectively, all P < 0.001). The Lumipulse G assay for Ab1-40 showed moderate correla- Within each CSF biomarker status, the proportion of pos- tion with values measured with Merck-Millipore’s ELISA itive amyloid scans varied when a second biomarker or (Pearson’s r of 0.76, P < 0.001). On the Lumipulse G ratio was taken into account. Figure S2 shows the 1818 ª 2019 The Authors. Annals of Clinical and Translational Neurology published by Wiley Periodicals, Inc on behalf of American Neurological Association. D. Alcolea et al. Amyloid PET and CSF Biomarkers in Lumipulse A Single biomarkers vs. 18F−Florbetapir Visual Status B Aβ1−42 ratios vs. 18F−Florbetapir Visual Status 1.0 1.0 0.8 0.8 Positive Percent Agreement Positive Percent Agreement 0.6 0.6 0.4 0.4 Aß1−42 AUC=0.76 ( 0.65 − 0.86) Aß1−40 Aß1−42/Aß1−40 AUC=0.59 (0.47 − 0.71) AUC=0.86 (0.77 − 0.96) 0.2 0.2 tTau tTau/Aß1−42 AUC=0.80 (0.70 − 0.89) AUC=0.87 (0.78 − 0.95) pTau pTau/Aß1−42 AUC=0.84 (0.75 − 0.93) AUC=0.88 (0.79 − 0.97) 0.0 0.0 1.0 0.8 0.6 0.4 0.2 0.0 1.0 0.8 0.6 0.4 0.2 0.0 Negative Percent Agreement Negative Percent Agreement Figure 1. Receiver operating characteristic analysis of individual (A) and combined (B) CSF biomarkers’ diagnostic utility to detect amyloid visual status. AUC, Area under the curve. proportion of positive amyloid scans within each combi- As seen in Figure 5, the agreement of CSF cutoffs with nation of two CSF biomarkers or ratios, and illustrates amyloid PET quantification was similar to that with the importance of considering a second biomarker. Of all visual classification for all biomarkers. We studied the participants with low CSF levels of Ab1-42, regardless of correlation between each CSF biomarker and global amy- the Ab1-42/Ab1-40 ratio status, 77% (56 out of 73) had a loid accumulation by fitting quadratic models. In these positive amyloid scan. This proportion increased to 87% models, the adjusted coefficients of determination (R2) (52 out of 60) within this group when the Ab1-42/Ab1- were higher for all ratios compared to individual 40 ratio was also low but decreased to 31% (four out of biomarkers. The combination of Ab1-42 with Ab1-40 13) when the Ab1-42/Ab1-40 ratio was high. In the group increased the R2 value to 0.44 (P < 0.001), indicating that of participants with high CSF levels of Ab1-42, the impact this ratio reflects the amyloid deposition better than Ab1- of considering the Ab1-42/Ab1-40 status had no effect, as 42 alone. Stratified analysis by 18F-Florbetapir visual sta- in all participants within this group the Ab1-42/Ab1-40 tus showed lower R2 values for all biomarkers, which sug- ratio was also high. These results highlight the importance gests that the correlation observed between CSF of using the Ab1-42/Ab1-40 ratio over Ab1-42 alone in biomarkers and amyloid PET quantification is partially the assessment of brain amyloidosis, especially in patients mediated by an amyloid-status effect. In this stratified with low CSF levels of Ab1-42. analysis, the highest correlation of SUVR values was found with the Ab1-42/Ab1-40 ratio in the amyloid–neg- ative group (R2 = 0.42; P < 0.001). Agreement of CSF cutoffs with amyloid quantification Discussion We next processed amyloid PET scans to obtain quantifi- cation values of amyloid deposition. In our study, the In our study, we determined cutoffs for four CSF previously validated SUVR value of 1.11,17 showed 83% biomarkers for AD (Ab1-42, Ab1-40, tTau and pTau) PPA, 76% NPA, and 81% OPA with visual classification. and their ratios measured on the fully automated LUMI- As displayed in Figure 4, scans that were divergently clas- PULSE G600II platform to optimize their concordance sified as “negative” or “positive” by one of the three with 18F-Florbetapir PET. We calibrated Ab1-42 levels to raters showed intermediate SUVR values compared to certified reference material, recently developed to harmo- scans that were unanimously classified. nize immunoassays across different platforms, and found ª 2019 The Authors. Annals of Clinical and Translational Neurology published by Wiley Periodicals, Inc on behalf of American Neurological Association. 1819 Amyloid PET and CSF Biomarkers in Lumipulse D. Alcolea et al. A B 1.00 1.00 Aβ1−42 = 916pg/ml 0.75 0.75 PPA=95%; NPA=51%; OPA=79% NPA NPA 0.50 0.50 PPA PPA Youden Youden 0.25 0.25 Aβ1−42 / Aβ1−40 = 0.062 PPA=88%; NPA=77%; OPA=84% 0.00 0.00 500 1000 1500 2000 0.03 0.06 0.09 0.12 Aβ1−42 (pg/ml) Aβ1−42 / Aβ1−40 C D 1.00 1.00 tTau = 456pg/ml 0.75 0.75 PPA=75%; NPA=83%; OPA=78% NPA NPA 0.50 0.50 PPA PPA Youden Youden 0.25 0.25 tTau / Aβ1−42 = 0.62 PPA=81%; NPA=83%; OPA=82% 0.00 0.00 500 1000 1500 0.0 0.5 1.0 1.5 2.0 tTau (pg/ml) tTau / Aβ1−42 E F 1.00 1.00 pTau = 63pg/ml 0.75 0.75 PPA=80%; NPA=83%; OPA=81% NPA NPA 0.50 0.50 PPA PPA Youden Youden 0.25 0.25 pTau / Aβ1−42 = 0.068 PPA=93%; NPA=80%; OPA=88% 0.00 0.00 100 200 300 0.0 0.1 0.2 0.3 0.4 pTau (pg/ml) pTau / Aβ1−42 Figure 2. Accuracy of all possible cutoff levels of individual (A, C, E) and combined (B, D, F) CSF biomarkers. Only those biomarkers that yielded areas under the curve above 0.70 and their ratios with Ab1-42 are shown. Vertical dotted lines indicate cutoffs with maximum Youden J index. PPA, Positive Percent Agreement; NPA, Negative Percent Agreement; OPA, Overall Percent Agreement. that the ratios Ab1-42/Ab1-40, tTau/Ab1-42 and pTau/ differ between these studies, and several methodological Ab1-42 had better diagnostic agreement with visual differences can explain these discrepancies. First, preana- assessment of amyloid scans than single biomarkers. As a lytical conditions, such as the type of collection and marker of amyloid pathology, the Ab1-42/Ab1-40 ratio storage tubes, are different between studies, and these had higher agreement with amyloid PET visual status and factors are known to have a great impact on the abso- showed better correlation with amyloid load quantifica- lute values of CSF biomarkers, especially for Ab1- tion compared to Ab1-42 alone. 42.18,19 Second, some analytical particularities for each The agreement between amyloid imaging and AD immunoassay and platform used in these studies (speci- CSF biomarkers has previously been studied by using ficity of the antibodies, time of incubation) result in other automated immunoassays.3,4,8 Our results are in diverse CSF biomarker measures. Calibration of all line with previous studies showing a good overall agree- automated platforms to certified reference material, cur- ment between amyloid imaging and AD CSF biomark- rently underway, will minimize this issue in the future. ers, higher for ratios than for single analytes.3,4 Likewise, differences in the affinity of PET radiotracers However, specific cutoff points for CSF biomarkers (11C-Pittsburg compound B, 18F-Flutemetamol or 1820 ª 2019 The Authors. Annals of Clinical and Translational Neurology published by Wiley Periodicals, Inc on behalf of American Neurological Association. D. Alcolea et al. Amyloid PET and CSF Biomarkers in Lumipulse A B C 2500 2500 PPA=75%; NPA=83%; OPA=78% PPA=80%; NPA=83%; OPA=81% 2500 tTau = 456pg/ml pTau = 63pg/ml 2000 2000 tTau / Aβ1−42 = 0.62 2000 pTau / Aβ1−42 = 0.068 Aβ1−42 / Aβ1−40 = 0.062 PPA=81%; NPA=83%; OPA=82% PPA=93%; NPA=80%; OPA=88% PPA=88%; NPA=77%; OPA=84% 1500 Aβ1−42 (pg/ml) 1500 Aβ1−42 (pg/ml) Aβ1−42 (pg/ml) 1500 1000 Aβ1−42 = 916pg/ml 1000 Aβ1−42 = 916pg/ml 1000 Aβ1−42 = 916pg/ml PPA=95%; NPA=51%; OPA=79% PPA=95%; NPA=51%; OPA=79% PPA=95%; NPA=51%; OPA=79% 500 500 500 0 0 0 5000 10000 15000 20000 25000 0 500 1000 1500 2000 0 100 200 300 Aβ1−40 (pg/ml) tTau (pg/ml) pTau (pg/ml) 18 18 18 F−Florbetapir Visual Status Positive Negative F−Florbetapir Visual Status Positive Negative F−Florbetapir Visual Status Positive Negative Figure 3. Agreement of visual amyloid status with single and combined CSF biomarkers. Panels A, B and C display scatterplots of CSF biomarker levels. Dashed lines indicate cutoffs that yielded maximum Youden J Index in the receiver operating characteristic analysis for each biomarker or ratio. PPA, Positive Percent Agreement; NPA, Negative Percent Agreement; OPA, Overall Percent Agreement. 18 F-Florbetapir) can lead to disparities in the selection study were selected by maximization of Youden J index. of cutoffs. Third, the composition of the populations This approach balances sensitivity and specificity and is was not the same across studies. Schindler et al. ana- equivalent to maximize accuracy for a pre-test disease lyzed data from community–dwelling volunteers,4 prevalence of 50%.20 However, other strategies might be whereas Janelidze et al. obtained their results from useful in certain clinical scenarios.21 For instance, for patients with mild cognitive impairment and subjective screening purposes, it might be helpful to apply cutoffs cognitive decline from the BioFINDER cohort.8 Hansson with high sensitivity, even when their specificity is et al. studied CSF of participants from ADNI and Bio- lower. For patients with clinically challenging diagnoses FINDER cohorts, that included cognitively normal vol- or in clinical trials, however, high specificity might be unteers, patients with mild cognitive impairment and preferable. Other possible approaches include the patients with AD dementia.3 In our study, we addition- sequential application of biomarker cutoffs. ally included patients with other dementias or neurode- The LUMIPULSE G600II has incorporated the possibil- generative diseases, which might reflect more realistically ity of measuring CSF levels of Ab1-40. In previous stud- the application of biomarkers in daily clinical practice. ies, and ours, this biomarker alone was not useful for the As in a number of other studies, the cutoffs in our detection of brain amyloid.4,7,9 Both Ab1-42 and the A B Rater 1 Rater 2 Rater 3 Global OPA = 81% OPA = 45% OPA = 100% OPA = 86% OPA=77% OPA=78% OPA=82% OPA=81% 1.75 1.75 1.50 1.50 18 F−Florbetapir 18 F−Florbetapir SUVR Visual Status SUVR Visual Status 1.25 Positive 1.25 Positive Negative Negative 1.00 1.00 0.75 0.75 Negative Positive Negative Positive Negative Positive Negative Positive 0/3 1/3 2/3 3/3 Number of raters that classified as 'Positive' Figure 4. Agreement between raters’ visual classification and amyloid quantification. OPA, Overall Percent Agreement; SUVR, Standardized Uptake Value Ratio. Panel A shows the agreement between amyloid quantification and rater’s individual and global visual assessments. Panel B shows the agreement between amyloid quantification and visual assessment stratified by the number of raters that assessed scans as positive. ª 2019 The Authors. Annals of Clinical and Translational Neurology published by Wiley Periodicals, Inc on behalf of American Neurological Association. 1821 Amyloid PET and CSF Biomarkers in Lumipulse D. Alcolea et al. A B 2500 All participants: R 2=0.22, p<0.001 All participants: R 2=0.44, p<0.001 Amyloid Positive: R 2=0.15, p=0.002 Amyloid Positive: R 2=0.07, p=0.031 Amyloid Negative: R 2=0.42, p<0.001 2000 Amyloid Negative: R 2=0.10, p=0.045 0.10 Aβ1−42 / Aβ1−40 Aβ1−42 (pg/ml) 18 18 F−Florbetapir F−Florbetapir 1500 Visual Status Visual Status Aβ1−42 / Aβ1−40 = 0.062 Positive Positive 1000 Aβ1−42 = 916pg/ml PPA=86%; PPA=86% NPA=77%; PA= PA OPA=83% Negative 0.05 Negative PPA=93%; A=9 NPA=50%; OPA=79% 500 0 0.00 0.75 1.00 1.25 1.50 1.75 0.75 1.00 1.25 1.50 1.75 SUVR SUVR C D 2000 All participants: R 2=0.27, p<0.001 All participants: R 2=0.41, p<0.001 Amyloid Positive: R 2=0.13, p=0.013 4 Amyloid Positive: R 2=0.22, p<0.001 Amyloid Negative: R 2=0.03, p= n.s. Amyloid Negative: R 2=0.10, p=0.080 1500 3 18 18 tTau / Aβ1−42 F−Florbetapir F−Florbetapir tTau (pg/ml) 1000 Visual Status Visual Status 2 Positive Positive Negative Negative 500 tTau tT Tau Ta a = 456pg/ml 6p p 1 tTau/Aβ1−42 tT 42 = 0.62 PPA=73%; NPA=83%; OPA=76% PPA=78%; NPA=80%; OPA=79% 0 0 0.75 1.00 1.25 1.50 1.75 0.75 1.00 1.25 1.50 1.75 SUVR SUVR E F 2 All participants: R =0.38, p<0.001 All participants: R 2=0.45, p<0.001 300 Amyloid Positive: R 2=0.15, p=0.004 Amyloid Positive: R 2=0.25, p<0.001 Amyloid Negative: R 2=0.21, p=0.012 Amyloid Negative: R 2=0.22, p=0.011 0.6 pTau / Aβ1−42 200 18 18 F−Florbetapir F−Florbetapir pTau (pg/ml) Visual Status Visual Status 0.3 Positive Positive 100 pTau pT T = 63pg/ml Ta 3p Negative pTau pT Tau / Aβ1−42 T 42 = 0.068 Negative PPA=81%; 81% NPA=83%; OPA=82% 0.0 PPA=88%; NPA=73%; OPA=83% 0 0.75 1.00 1.25 1.50 1.75 0.75 1.00 1.25 1.50 1.75 SUVR SUVR Figure 5. Scatterplots and correlations of amyloid quantification values with individual biomarkers (A, C, E) and ratios (B, D, F). Correlation between SUVR values and CSF biomarkers was assessed by fitting quadratic models for all participants (black) and after stratifying by visual amyloid status (red and green). Shaded areas indicate 95% confidence intervals. Dashed vertical lines indicate the SUVR cutoff of 1.11 as in Landau et al. Horizontal lines correspond to cutoffs for each CSF biomarker and ratio. PPA, NPA and OPA values correspond to the agreement between amyloid quantification and CSF biomarkers. PPA, Positive Percent Agreement; NPA, Negative Percent Agreement; OPA, Overall Percent Agreement; SUVR, Standardized Uptake Value Ratio. Ab1-42/Ab1-40 ratio are included in the “A” category of quantification. Our results are in line with previous litera- the ATN classification system together with amyloid PET, ture that suggests that the use of the Ab1-42/Ab1-40 ratio but in line with other studies,7,9,22,23 we found that the could compensate individual differences in amyloid pre- Ab1-42/Ab1-40 ratio had better agreement with visual cursor protein processing that otherwise might lead to amyloid status and higher correlation with brain amyloid false positive or false negative Ab1-42 CSF levels.24,25 This 1822 ª 2019 The Authors. Annals of Clinical and Translational Neurology published by Wiley Periodicals, Inc on behalf of American Neurological Association. D. Alcolea et al. Amyloid PET and CSF Biomarkers in Lumipulse information adds to the fact that using the Ab1-42/Ab1- Acknowledgment 40 ratio has proven to partially mitigate the effect of some preanalytical confounders that have been described We are grateful to all participants in the study and to all to alter the results of amyloid levels.26–28 Altogether, our the members of the clinical team that were involved in data suggest that the use of the Ab1-42/Ab1-40 ratio the selection and assessment of participants. would be more reliable in clinical practice than Ab1-42 alone as a marker of amyloidosis and that this combina- Conflict of Interest tion should be used in routine. The main strength of our study is that four AD CSF D.A. participated in advisory boards from Fujirebio-Eur- biomarkers (Ab1-42, Ab1-40, tTau and pTau) measured ope and received speaker honoraria from Fujirebio-Eur- simultaneously with the automated Lumipulse G System ope, Nutricia and from Krka Farmaceutica S.L. R.B. were compared for the first time to 18F-Florbetapir PET participated in advisory boards from Lilly and Nutricia, to calculate amyloid–based cutoffs. In addition, this is, to and he received speaker honoraria and travel funding our knowledge, the first study to present Ab1-42 levels from Novartis and Nutricia. A.L. participated in advisory that have been standardized to certified reference mate- boards from Fujirebio-Europe, Nutricia, Biogen, and rial, recently developed to harmonize immunoassays received speaker honoraria from Lilly. N.LB. and E.H. are across different platforms. Standardized values will make employed by Fujirebio Europe. N.V. A.N. and V.O. are our study more easily comparable to future studies. employed by Fujirebio Iberia, S.L. Moreover, to avoid possible sources of variability, we fol- lowed homogeneous CSF preanalytical and analytical pro- References cedures and used the same batch of reagents for all 1. Jack CR, Bennett DA, Blennow K, et al. NIA-AA Research measurements. Also, the inclusion of participants with Framework: toward a biological definition of Alzheimer’s neurodegenerative diseases outside the AD spectrum pro- disease. Alzheimer’s Dement 2018;14:535–562. vides a more realistic application of biomarkers in daily 2. Toledo JB, Bjerke M, Da X, et al. Nonlinear association clinical practice. between cerebrospinal fluid and florbetapir F-18 b-amyloid However, we are aware of some limitations. We did not measures across the spectrum of Alzheimer disease. JAMA test the effect that deviations from our preanalytical proto- Neurol 2015;4283:1–11. col would have on the final cutoffs, and therefore, the cut- 3. Hansson O, Seibyl J, Stomrud E, et al. CSF biomarkers of offs that we report should be taken cautiously under other Alzheimer’s disease concord with amyloid-b PET and operating procedures. Some of the clinical categories in predict clinical progression: a study of fully automated our study included a small number of participants. immunoassays in BioFINDER and ADNI cohorts. Although we used amyloid positivity/negativity as the gold Alzheimer’s Dement 2018;14:1470–1481. standard in our study, the composition of the sample in 4. Schindler SE, Gray JD, Gordon BA, et al. Cerebrospinal terms of clinical diagnosis might be relevant for the inter- fluid biomarkers measured by Elecsys assays compared to pretation and applicability of our results. Additionally, amyloid imaging. Alzheimer’s Dement 2018;14;1460–1469. only very few participants had additional Tau imaging 5. Fagan AM, Head D, Shah AR, et al. Decreased and/or 18F-Fluorodeoxyglucose-PET, and we could not cerebrospinal fluid Abeta(42) correlates with brain atrophy compare the agreement of CSF pTau and tTau with in cognitively normal elderly. Ann Neurol 2009;65:176–183. molecular imaging markers of the “T” and the “N” cate- 6. Illan-Gala I, Pegueroles J, Montal V, et al. Challenges gories of the ATN classification. Likewise, as participants associated with biomarker-based classification systems for in this study are part of a living cohort, neuropathological Alzheimer’s disease. Alzheimer’s Dement 2018;10:346–357. confirmation is not available at this moment. 7. Janelidze S, Zetterberg H, Mattsson N, et al. CSF Ab42/ In this study, we found that the Ab1-42 ratios to Ab1- Ab40 and Ab42/Ab38 ratios: better diagnostic markers of 40, tTau and pTau in CSF show a good agreement with Alzheimer disease. Ann Clin Transl Neurol 2016;3:154– amyloid visual status and that the Ab1-42/Ab1-40 ratio 165. had better correlation with the amount of amyloid bur- 8. Janelidze S, Pannee J, Mikulskis A, et al. Concordance den compared to Ab1-42 alone. The understanding of the between different amyloid immunoassays and visual agreement between CSF biomarkers and amyloid imaging amyloid positron emission tomographic assessment. JAMA is crucial to identify situations in which these two modal- Neurol 2017;74:1492–1501. ities might not be interchangeable. This information has 9. Lewczuk P, Matzen A, Blennow K, et al. Cerebrospinal to be taken into consideration both in the diagnostic fluid Ab42/40 corresponds better than Ab42 to amyloid assessment in clinical practice and in the selection of par- PET in Alzheimer’s disease. J Alzheimers Dis 2017;55: ticipants in clinical trials. 813–822. ª 2019 The Authors. Annals of Clinical and Translational Neurology published by Wiley Periodicals, Inc on behalf of American Neurological Association. 1823 Amyloid PET and CSF Biomarkers in Lumipulse D. Alcolea et al. 10. Kuhlmann J, Andreasson U, Pannee J, et al. CSF Ab 1–42 – 21. Lle o A, Irwin DJ, Illan-Gala I, et al. A 2-step cerebrospinal an excellent but complicated Alzheimer’s biomarker – a algorithm for the selection of frontotemporal lobar route to standardisation. Clin Chim Acta 2017;467:27–33. degeneration subtypes. JAMA Neurol 2018;75:738–745. 11. Alcolea D, Carmona-Iragui M, Suarez-Calvet M, et al. 22. Pannee J, Portelius E, Minthon L, et al. Reference Relationship between b-secretase, inflammation and core measurement procedure for CSF amyloid beta (Ab)1–42 cerebrospinal fluid biomarkers for Alzheimer’s disease. J and the CSF Ab1-42/Ab1-40 ratio - a cross-validation study Alzheimers Dis 2014;42:157–167. against amyloid PET. J. Neurochem 2016;139:651–658. 12. Alcolea D, Vilaplana E, Pegueroles J, et al. Relationship 23. Niemantsverdriet E, Ottoy J, Somers C, et al. The between cortical thickness and cerebrospinal fluid YKL-40 cerebrospinal fluid AB1–42/AB1–40 ratio improves in predementia stages of Alzheimer’s disease. Neurobiol concordance with amyloid-PET for diagnosing Alzheimer’s Aging 2015;36:2018–2023. disease in a clinical setting. J Alzheimer’s Dis 2017;60:561– 13. Carmona-Iragui M, Fernandez-Arcos A, Alcolea D, et al. 576. Cerebrospinal fluid anti-amyloid-b autoantibodies and 24. Wiltfang J, Esselmann H, Bibl M, et al. Amyloid beta amyloid PET in cerebral amyloid angiopathy-related peptide ratio 42/40 but not A beta 42 correlates with inflammation. J Alzheimer’s Dis 2015;50:1–7. phospho-Tau in patients with low- and high-CSF A beta 14. Alcolea D, Vilaplana E, Suarez-Calvet M, et al. CSF 40 load. J Neurochem 2007;101:1053–1059. sAPPb, YKL-40, and neurofilament light in 25. Dumurgier J, Schraen S, Gabelle A, et al. Cerebrospinal frontotemporal lobar degeneration. Neurology fluid amyloid-b 42/40 ratio in clinical setting of memory 2017;89:178–188. centers: a multicentric study. Alzheimer’s Res Ther 2015;7. 15. Carmona-Iragui M, Balasa M, Benejam B, et al. Cerebral 26. Gervaise-Henry C, Watfa G, Albuisson E, et al. amyloid angiopathy in Down syndrome and sporadic and Cerebrospinal fluid Ab42/Ab40 as a means to limiting autosomal-dominant Alzheimer’s disease. Alzheimer’s tube- and storage-dependent pre-analytical variability in Dement 2017;13:1251–1260. clinical setting. J Alzheimers Dis 2017;57:437–445. 16. Rieck JR, Rodrigue KM, Kennedy KM, et al. The effect of 27. Toombs J, Foiani MS, Wellington H, et al. Amyloid b beta-amyloid on face processing in young and old adults: peptides are differentially vulnerable to preanalytical a multivariate analysis of the BOLD signal. Hum Brain surface exposure, an effect incompletely mitigated by the Mapp 2015;36:2514–2526. use of ratios. Alzheimer’s Dement 2018;10:311–321. 17. Landau SM, Mintun MA, Joshi AD, et al. Amyloid 28. Delaby C, Mu~ noz L, Torres S, et al. Impact of CSF storage deposition, hypometabolism, and longitudinal cognitive volume on the analysis of Alzheimer’s disease biomarkers on decline. Ann Neurol 2012;72:578–586. an automated platform. Clin Chim Acta 2019;490:98–101. 18. Bjerke M, Portelius E, Minthon L, et al. Confounding factors influencing amyloid Beta concentration in cerebrospinal fluid. Int J Alzheimers Dis 2010;2010:1–11. Supporting Information 19. Hansson O, Mikulskis A, Fagan AM, et al. The impact of Additional supporting information may be found online preanalytical variables on measuring cerebrospinal fluid in the Supporting Information section at the end of the biomarkers for Alzheimer’s disease diagnosis: a review. article. Alzheimer’s Dement 2018;14:1313–1333. 20. Bartlett JW, Frost C, Mattsson N, et al. Determining Figure S1. Interassay coefficients of Ab1-42 (A), Ab1-40 cut-points for Alzheimer’s disease biomarkers: statistical (B), tTau (C) and pTau (D). issues, methods and challenges. Biomark Med Figure S2. Agreement of visual amyloid PET with the 2012;6:391–400. combination of two CSF biomarkers or ratios. 1824 ª 2019 The Authors. Annals of Clinical and Translational Neurology published by Wiley Periodicals, Inc on behalf of American Neurological Association.