**3. Variability introduced by the analytical step**

There are several available assays for the determination of CSF Aβ42, T-Tau and P-Tau, commercialized by different companies (Covance, Cusabio, IBL international, Innogenetics, Invitrogen, Millipore, Meso Scale Discovery, Wako... list not exhaustive). Large variation, in assay performance and outcomes of CSF Aβ42, T-Tau and P-Tau levels was observed between laboratories also when the same assay format was used, reaching in some cases an inter-assay and inter-laboratory coefficient variations of 20 to 35% [7, 36]. As shown in conclusions of the first report of the external quality control (EQC) program started by the Alzheimer's associa‐ tion [37], ELISA techniques dominate the market while multiplex techniques are used less. In this program, for Aβ42, T-Tau and P-Tau, most of laboratories [26 laboratories) used the INNOTEST enzyme-linked immunosorbent assays (ELISAs) (Innogenetics, Ghent, Belgium, www.innogenetics.com), whereas 14 laboratories used the bead-based Luminex xMAP platform with the INNO-BIA AlzBio3 (Innogenetics, Ghent, Belgium, www.innogenet‐ ics.com). Moreover, for Aβ42 and T-Tau, 5 laboratories used Meso Scale Discovery (MSD, Gaithersburg, MD, www.mesoscale.com) technology [37].

**3.2. Extent of the variability highlighted by this EQC program [37]**

In this report, results were grouped according to analytical techniques and samples [37]. The total CVs among centers were 16% to 28% for ELISA, 13% to 36% for xMAP, and 16% to 36% for MSD. CVs for MSD must be interpreted with caution, because they included 2 different Monoclonal antibodies (Mab) for Aβ<sup>42</sup> assays, binding to different epitopes on the amyloid peptide. These data were totally conformed to those reported earlier [7, 36]. There was no major modification of the CV in the longitudinal evaluation, except a decrease in variation for T-tau measured by ELISA. This was expected, since there was no active

Pre-Analytical and Analytical Critical Factors Influencing the High Variability of the Concentrations Levels of...

http://dx.doi.org/10.5772/55512

185

Within-laboratory CVs were examined at the reference laboratories for ELISA and xMAP in two consecutive rounds. CVs were 3.2% to 24% for ELISA and 2.3% to 26% for xMAP, but differed between analytes within individual laboratories, indicating assay-dependent

The analytical techniques reported different absolute values for the biomarkers. ELISA values for Aβ42, were about 2 fold higher than xMAP values. MSD values for Aβ42, were dependent of the Mab used. ELISA values for T-Tau were about 3 fold higher than xMAP values. Finally for P-Tau, the differences inter techniques were clearly decreased in comparison to Aβ42, and T-Tau. Considerable variability exists among the same manufacturer between mono and multiplex technology. For example, the decision threshold of clinical disease was reported to be at 86 pg / mL and 350 pg / mL for T-TAU measured by xMAP technology of Innogenetics on the platform Luminex and the conventional ELISA, respectively [39]. Factors of correction between values obtained by xMAP and ELISA, were used for global comparison of groups of patients, i.e. controls, Mild Cognitive Impairment and AD patients to predict incipient AD by CSF biomarkers [40]. In an other side, it was clearly shown that the use of factors of correction did not resolve the discrepancy in values observed between xMAP and ELISAs [41]. Although the observed biomarker concentrations may vary significantly between platforms, including MSD, xMAP and ELISA, these techniques seem to have similar diagnostic accuracy for patients

In this study analysing the variability of results from only two rounds of an EQC program and from many different assay lots used, the authors limited their interpretation of the relative contributions from between-laboratory, within-laboratory, and between-lot components to the total variability [37]. Differences in within-laboratory CVs among the biomarkers within individual reference laboratories suggest that assay-related factors are important. Moreover,

with AD versus controls [39] or for detecting early AD [41, 42].

*3.2.1. Total variability*

intervention between the 2 rounds [37].

*3.2.2. Within-laboratory precision*

*3.2.3. Differences in absolute values*

**3.3. Possible sources of variability**

variations [37].

#### **3.1. Principles of assays**

INNOTEST enzyme-linked immunosorbent assays (ELISAs) (Innogenetics) are classical ELISAs with colorimetric detection.

INNO-BIA AlzBio3 allows the simultaneous quantification of Aβ42, T-Tau and P-Tau in CSF using xMAP® technology (xMAP is a registered trademark of Luminex Corp). The micro‐ sphere-based Luminex xMAP technology involves covalent coupling of a capture antibody to spectrally specific fluorescent microspheres [38]. Each microsphere number has a unique spectral identity. The classification of each bead is made by excitation at 635 nm. Each bead number is linked with only one antibody and the signals from analytes in the mixture are identified unequivocally. The quantification of the molecular reaction that has occurred at the microsphere surface, is done using a fluorochrome, the phycoerythrin coupled to streptavidin. The intensity of the fluorescence, derived after excitation of PE at 532 nm, is reported.

MSD offers the possibility to measure in simplex or multiplex format, depending on the biomarker analysed. Whereas t-Tau is measured in simplex format by the participants of the external control program, Aβ42 can be measured in simplex or multiplex format in combination with Aβ<sup>38</sup> and Aβ40. Multi-array plate formats include 96- and 384-well plates. The multi-spot plates are available with up to 100 spots per well. MSD uses electrochemi‐ luminescence to detect binding events on patterned arrays. Electrochemiluminescence detection uses labels that emit light at ~620 nm when electrochemically stimulated, the stimulation mechanism (electricity) being decoupled from the signal (light). The signals are treated by the SECTOR Imager Instrument, which is medium throughput imaging detection systems (charge-coupled device camera), capable of multiplexing in all spot formats and reads 96- and 384-well plates.

### **3.2. Extent of the variability highlighted by this EQC program [37]**

#### *3.2.1. Total variability*

**3. Variability introduced by the analytical step**

Gaithersburg, MD, www.mesoscale.com) technology [37].

**3.1. Principles of assays**

184 Understanding Alzheimer's Disease

ELISAs with colorimetric detection.

reads 96- and 384-well plates.

There are several available assays for the determination of CSF Aβ42, T-Tau and P-Tau, commercialized by different companies (Covance, Cusabio, IBL international, Innogenetics, Invitrogen, Millipore, Meso Scale Discovery, Wako... list not exhaustive). Large variation, in assay performance and outcomes of CSF Aβ42, T-Tau and P-Tau levels was observed between laboratories also when the same assay format was used, reaching in some cases an inter-assay and inter-laboratory coefficient variations of 20 to 35% [7, 36]. As shown in conclusions of the first report of the external quality control (EQC) program started by the Alzheimer's associa‐ tion [37], ELISA techniques dominate the market while multiplex techniques are used less. In this program, for Aβ42, T-Tau and P-Tau, most of laboratories [26 laboratories) used the INNOTEST enzyme-linked immunosorbent assays (ELISAs) (Innogenetics, Ghent, Belgium, www.innogenetics.com), whereas 14 laboratories used the bead-based Luminex xMAP platform with the INNO-BIA AlzBio3 (Innogenetics, Ghent, Belgium, www.innogenet‐ ics.com). Moreover, for Aβ42 and T-Tau, 5 laboratories used Meso Scale Discovery (MSD,

INNOTEST enzyme-linked immunosorbent assays (ELISAs) (Innogenetics) are classical

INNO-BIA AlzBio3 allows the simultaneous quantification of Aβ42, T-Tau and P-Tau in CSF using xMAP® technology (xMAP is a registered trademark of Luminex Corp). The micro‐ sphere-based Luminex xMAP technology involves covalent coupling of a capture antibody to spectrally specific fluorescent microspheres [38]. Each microsphere number has a unique spectral identity. The classification of each bead is made by excitation at 635 nm. Each bead number is linked with only one antibody and the signals from analytes in the mixture are identified unequivocally. The quantification of the molecular reaction that has occurred at the microsphere surface, is done using a fluorochrome, the phycoerythrin coupled to streptavidin.

The intensity of the fluorescence, derived after excitation of PE at 532 nm, is reported.

MSD offers the possibility to measure in simplex or multiplex format, depending on the biomarker analysed. Whereas t-Tau is measured in simplex format by the participants of the external control program, Aβ42 can be measured in simplex or multiplex format in combination with Aβ<sup>38</sup> and Aβ40. Multi-array plate formats include 96- and 384-well plates. The multi-spot plates are available with up to 100 spots per well. MSD uses electrochemi‐ luminescence to detect binding events on patterned arrays. Electrochemiluminescence detection uses labels that emit light at ~620 nm when electrochemically stimulated, the stimulation mechanism (electricity) being decoupled from the signal (light). The signals are treated by the SECTOR Imager Instrument, which is medium throughput imaging detection systems (charge-coupled device camera), capable of multiplexing in all spot formats and In this report, results were grouped according to analytical techniques and samples [37]. The total CVs among centers were 16% to 28% for ELISA, 13% to 36% for xMAP, and 16% to 36% for MSD. CVs for MSD must be interpreted with caution, because they included 2 different Monoclonal antibodies (Mab) for Aβ<sup>42</sup> assays, binding to different epitopes on the amyloid peptide. These data were totally conformed to those reported earlier [7, 36]. There was no major modification of the CV in the longitudinal evaluation, except a decrease in variation for T-tau measured by ELISA. This was expected, since there was no active intervention between the 2 rounds [37].

#### *3.2.2. Within-laboratory precision*

Within-laboratory CVs were examined at the reference laboratories for ELISA and xMAP in two consecutive rounds. CVs were 3.2% to 24% for ELISA and 2.3% to 26% for xMAP, but differed between analytes within individual laboratories, indicating assay-dependent variations [37].

#### *3.2.3. Differences in absolute values*

The analytical techniques reported different absolute values for the biomarkers. ELISA values for Aβ42, were about 2 fold higher than xMAP values. MSD values for Aβ42, were dependent of the Mab used. ELISA values for T-Tau were about 3 fold higher than xMAP values. Finally for P-Tau, the differences inter techniques were clearly decreased in comparison to Aβ42, and T-Tau. Considerable variability exists among the same manufacturer between mono and multiplex technology. For example, the decision threshold of clinical disease was reported to be at 86 pg / mL and 350 pg / mL for T-TAU measured by xMAP technology of Innogenetics on the platform Luminex and the conventional ELISA, respectively [39]. Factors of correction between values obtained by xMAP and ELISA, were used for global comparison of groups of patients, i.e. controls, Mild Cognitive Impairment and AD patients to predict incipient AD by CSF biomarkers [40]. In an other side, it was clearly shown that the use of factors of correction did not resolve the discrepancy in values observed between xMAP and ELISAs [41]. Although the observed biomarker concentrations may vary significantly between platforms, including MSD, xMAP and ELISA, these techniques seem to have similar diagnostic accuracy for patients with AD versus controls [39] or for detecting early AD [41, 42].

#### **3.3. Possible sources of variability**

In this study analysing the variability of results from only two rounds of an EQC program and from many different assay lots used, the authors limited their interpretation of the relative contributions from between-laboratory, within-laboratory, and between-lot components to the total variability [37]. Differences in within-laboratory CVs among the biomarkers within individual reference laboratories suggest that assay-related factors are important. Moreover, the high variability of the results of biomarkers measured by different commercial kits can be explained, by the use of different antibodies, the nature of the calibrator, the calibration method and many others factors as for example the nature of standard. Increasing data during years and by incorporation of new centers (since this first report concerning 40 laboratories, in summer of 2012, 64 laboratories were participating at this program) will permit to better identify the major sources of variability in analytical steps. Thus, we can just list the different points to be further investigated.

sample (native CSF pools, spiked CSF with standards, peptides...) how many QC samples, range of concentrations to cover, absence of reference material. This point is crucial for laboratories concerned by accreditation scheme based on the application of ISO15189 standard.

Pre-Analytical and Analytical Critical Factors Influencing the High Variability of the Concentrations Levels of...

http://dx.doi.org/10.5772/55512

187

Many crucial points need to be solved as the poor quality of the test procedure instructions to decrease variability induced by misunderstanding of the protocols. This lacking information is often an indicator of minimal method optimization of the protocol (for example incubation steps, handling the reagents...). The reagents must be proposed in a manner that permits to decrease variability, for instance the «ready to use» calibrators. The absence of quality control included in the kit is a major problem. In fact, part of the discrepancy observed in the concen‐ trations levels between the analytical techniques ELISA, xMAP, and MSD is caused by the lack of certified reference materials (CRMs). This could mainly impact the interlot variability and is at least, a brake to standardization. Antibody purification, coating of plates and beads are

The present chapter highlights two main issues responsible for the lack of harmonization of CSF AD biomarkers cut-offs values: the lack of standardization of the pre-analytical steps and the high variability of results linked to the analytical step. This latter issue can be explained by the absence of transferability of results between the different platforms but also by the high inter laboratory dispersion within the same assays. Previous consensus guidelines for preanalytical factor standardization gave the way to resolve this issue, evidencing the need to standardize sampling and storage tubes, the type of the needle for the CSF puncture and the long term storage. Establishing SOPs for sample processing would allow to compare diagnostic conclusions between different laboratories. The implementation of those SOPs in the clinical community may reduce part of the variability found in the analysis of AD CSF biomarkers. Antibody purification, coating surfaces, preparation of standards, manufacturers instructions are also sources of variation, which need to be decreased and requires increased efforts by kit manufacturers. The optimal approach is a collaborative effort between commercial kit and instrument platform manufacturers, laboratories concerned by those methods, and reference

We wish to thank all our collaborators of the JPND BIOMARKPD program, those of the French Society of Clinical Biology (SFBC) and those of the NEUROSCREEN European project for their

*3.3.1. Issues to be solved by manufacturers*

also factors of lot-lot variability.

**4. Conclusion**

standardization programs.

**Acknowledgements**

valuable assistance.

#### **In the laboratory, the biologist will take care for:**

**a.** Pipetting

The pipetting mode (inverse pipetting...) is not specified by the manufacturer. Using a single tip can influence the standard curve accuracy. However, the magnitude of this effect, if any, should be tested, to provide a better basis for recommendation [43].

**b.** Calibration

For lyophilised standard, accurate solubilization and accurate pipetting is critical. Moreover, since for INNOTEST Ab42, the first point of the curve calibration must be adjusted depending of the set value, accurate pipeting is absolutely needed. The type of curve fitting used and the software for data calculation were shown as possible factors of variability [43].

**c.** Reagent handling and adhesion of biologists to the manufacturer standard operating procedure (SOP)

The adhesion of routine laboratories to the manufacturer SOP is absolutely needed to reduce the part of the variability found in CSF biomarkers analysis. For that, a great effort must be done by the different manufacturers to limit individual interpretation of the technical instruc‐ tions. The best example consists in the definition of the «room temperature» which can mainly vary from the north to the south of Europe. The maintenance of laboratory equipment is a crucial point to ensure the accuracy of pipeting volumes, the accuracy of temperatures, the accuracy of detection signals and the quality and reproducibility of washing steps.

**d.** Familiarization with the method and Competency Train

Implementing these techniques in the laboratory needs a training program ensured by the manufacturer. Moreover, habilitation and qualification of the laboratory staff must be done.

**e.** Validation criteria of runs for rejecting data

Different means are used to ensure validation of results. The definition of the criteria of acceptance of results must be strict. They include the calibration curve parameters, the CV of the duplicate samples and the use of an internal quality control program. For the CV criteria acceptance, in our experience, it seems that they are to be adequately defined since, the recommendation of CV < 20% done in the INNOTEST documentation, is not acceptable all along the dynamical range of the assay, in particular when the concentration level is near the clinical cut-off. Moreover, in the absence of QC samples in the kit, the biologist needs to implement its own QC program with different crucial points to resolve: the nature of the sample (native CSF pools, spiked CSF with standards, peptides...) how many QC samples, range of concentrations to cover, absence of reference material. This point is crucial for laboratories concerned by accreditation scheme based on the application of ISO15189 standard.

#### *3.3.1. Issues to be solved by manufacturers*

Many crucial points need to be solved as the poor quality of the test procedure instructions to decrease variability induced by misunderstanding of the protocols. This lacking information is often an indicator of minimal method optimization of the protocol (for example incubation steps, handling the reagents...). The reagents must be proposed in a manner that permits to decrease variability, for instance the «ready to use» calibrators. The absence of quality control included in the kit is a major problem. In fact, part of the discrepancy observed in the concen‐ trations levels between the analytical techniques ELISA, xMAP, and MSD is caused by the lack of certified reference materials (CRMs). This could mainly impact the interlot variability and is at least, a brake to standardization. Antibody purification, coating of plates and beads are also factors of lot-lot variability.
