The Alzheimer’s Association QC program for CSF biomarkers • Ongoing ( 2009-) project to monitor CSF biomarker lab and assay performance • Led by Gothenburg University, funded by the Alzheimer’s Association Principle for the QC program: For each round, 3 QC samples (pooled CSF) are sent out 2 unique samples - for comparisons between labs 1 identical sample - for comparisons over time Frequency: 3 times per year > 90 labs At present 17 rounds completed Newly added assay developments and formats: - Innotest ELISAs with ready-to-use (RTU) calibrators - MSD 6E10 Vplex assays - EUROIMMUNE/ADx assays - Roche cobas Elecsys fully automated assay

First paper in 2011 – Rounds 1-2  Between laboratory CVs of 13-36%

Results from Rounds 7-17 - Innotest Between laboratory CVs for the Innotest assays tend to go down No clear differences between Aβ42 and T-tau, but P-tau show somewhat lower variability  Aliquots of pooled CSF are stable over >3.5 years !

Results from Rounds 7-17 – AlzBio3 Between laboratory CVs for AlzBio3 show no clear trend for reduced variability  Few labs make results more uncertain

Results from Rounds 7-17 – Meso Scale (MSD) Between laboratory CVs for MSD(6E10) Aβ assays show a tendency for reduced variability

Results from Rounds 7-17 – EuroImmune / ADx 16A 16B 17A 17B Aβ42 57 % 19.4 % 6.5 % 8.2 % Aβ40 45.5 % 22.7 % 9.1% 8.4 % T-tau 6% 21 % 17.4% - EuroImmune / ADx results on the A and B samples The EuroImmune / ADx asays have just been enrolled in the QC program  Few labs and few rounds make results uncertain – last round shows good performance for the Aβ assays

Sources of variability for CSF biomarker measurements CAUSE Pre-analytical factors • E.g. type of test tubes - Not applicable in the QC program Analytical factors • Laboratory standardization (e.g. pipette calibration) • Analytical procedures (e.g. time, temperature, manual or pipetting robot) • Training / certification of technicians (e.g. pipetting variability) • Data handling (e.g. calibration curve) Assay manufacturing • Reagent (e.g. antibody) purity and variability • Kit production (e.g. plate coating) • Calibrator stability and format • Lot-to-lot consistency (batch bridging procedure)  Analytical factors and assay quality and stability have a large impact on outcome

Sources of variability for CSF biomarker in the QC program INNOTEST ELISA Between lot Between laboratory Contribution to variation % Overall variability CV % 23% 18% 19%  Analytical factors (laboratory and technician procedures) and assay quality and stability have a large impact on variability Mattson N et al. Alzheimer’s & Dementia 2013;9: 251-261

Impact of automation of ELISA methods Data from the Clinical Neurochemistry Lab, Sahlgrenska University Hospital, Sweden Internal quality control (QC) samples (pooled CSF) run on every ELISA plate (Innotest) All runs during 1 year Manual ELISA 134 QC samples – CV 14.2% Pipetting robot - Tecan 230 QC samples – CV 11.0% Automation of ELISAs for CSF Aβ(1–42) provides some improvement in variability

CSF biomarker assays on fully automated clinical analyzers • Fully automated lab analyzer – no variation due to differences in laboratory procedures or between technicians – will give reduced between-run, between-batch and between-lab variations  Paves the way for uniform cut-off levels Single sample analysis  no need to ”wait” for samples to fill up an ELISA plate  fast results (< 30 min) to the clinician Roche Diagnostics – Cobas Elecsys

Roche Diagnostics Elecsys assay – first Round 2016 14A 14B 15A 15B 16A 16B 17A 17B Aβ42 2.9 % 4.4 % 4.6 % 3.4 % 3.0 % 2.5 % 1.9 % 3.2 % Cobas Elecsys results on the A and B samples  The Cobas Elecsys assay show minimal variability

The CSF biomarkers Aβ42 and T-tau on the fully automated Lumipulse instruments Characteristics: Lumipulse G1200 Lumipulse G600 II (benchtop model) Reaction time 25 minutes 25 minutes Throughput 120 tests per hour 60 tests per hour Analytical performance (in-house data) CSF Aβ42  CVs of 2-5% within-run, between instrument and between-day CSF T-tau  CVs of 1-3% within-run, between instrument and between-day The Lumipulse assays are not yet in the QC program

The IFCC Work Group for CSF proteins and The Global Biomarker Standardization Consortium High precision Mass spectrometry-based technique Certified calibrator (amino acid analysis) Tested in Round Robin studies Intended use Set level in the Certified Reference Material Reference Measurement Procedure (RMP) Golden standard method for absolute quantification Large aliquoted CSF pool Exact level set using RMPs Commutability between assays tested Tested for long-term stability etc. Intended use Distribution to kit vendors and large labs Harmonize CSF levels between assay formats Assure stability between production lots Certified Reference Material (CRM) Golden standard CSF with exact levels

Reference method for CSF Aβ42 - Validated ”Golden standard” method Antibody-free Single Reaction Monitoring (SRM) Triple Quad mass spec method for CSF Aβ isoforms Separation of Aβ by HPLC Vantage Triple Quad mass spectrometry CSF + internal Aβ standards Guanidine HCl  denaturation SPE antibody-free purification Elute + dry + redissolve in ACN and NH4OH Quantification of Aβ isoforms 5.5 6.0 6.5 7.0 7.5 Time (min) Aβ1-38 Aβ1-42 Aβ1-40 Endogenous Aβ Aβ standards • Isotope labelled Aβ calibrator added to the CSF sample (and thus processed identically) • No antibodies involved  absolute quantification without interference (matrix effects)

Candidate reference methods for CSF Aβ42 published Mean of all labs Value for Individual lab • Four laboratories with SRM methods for CSF Aβ42: University of Gothenburg University of Pennsylvania PPD, USA Waters, USA • 12 CSF samples analyzed • One CSF served as a Candidate Reference Material  SRM mass spec suitable as a Reference Measurement Procedure (RMP) for CSF Aβ42

How do amyloid PET and CSF Aβ42 compare ? Study design: 118 patients with cognitive complaint examined for both CSF biomarkers - as part of clinical routine – 2 years and amyloid 18F-flutemetamol PET Cut-offs: CSF Aβ42 < 647 pg/mL 18F-flutemetamol PET > 1.42 Original cohort n= 118 Positive PET+CSF or Negative PET+CSF 92 % Validation cohort n= 38 Positive PET+CSF or Negative PET+CSF 97 % Palmquist S, et al, JAMA Neurol 2014

Mass spec method for CSF Aβ42 vs. amyloid PET Study design: 103 patients with cognitive complaint examined for CSF Aβ using SRM mass spec and amyloid 18F-flutemetamol PET Aβ42/40 ratio Aβ42  High concordance between amyloid PET and SRM mass spec, especially for the CSF Aβ42/40 ratio Shahim P, et al, unpublished

How can we apply absolute biomarker cut-off levels for AD ? Younger AD patients showed extensive pathological changes (cortical plaques and tangles) when compared with people of similar age, but in older age, pathological changes in AD patients approached that of old age alone.  Marked overlap in pathology (plaques and tangles) between AD and aging  Elderly demented patients have many different types of pathology (not only plaques and tangles)  Even if we have exact biomarkers – how do we apply a cut-off ?

“Absolute levels of CSF Aβ42” vs. “amyloid PET positive/negative” Cut-off <192 pg/mL A Normal Low Cut-off >1.11 SUVR B Positive Negative  Continuum in pathology between AD and aging gives a continuum in CSF Aβ levels and amyloid PET SUVs  Is it better to report actual values to the clinician instead of “positive” or “negative” ?

We are doing great progress in the GBSC and IFCC WG-CSF - Standardization may be difficult