Abstract
Background
Different cerebrospinal fluid (CSF) amyloid-beta 1–42 (Aβ 1–42), total Tau (Tau) and Tau phosphorylated at threonine 181 (P-Tau) levels are reported, but currently there is a lack of quality control programmes. The aim of this study was to compare the measurements of these CSF biomarkers, between and within centres.
Methods
Three CSF-pool samples were distributed to 13 laboratories in 2004 and the same samples were again distributed to 18 laboratories in 2008. In 2004 six laboratories measured Aβ 1–42, Tau and P-Tau and seven laboratories measured one or two of these marker(s) by enzyme-linked immunosorbent assays (ELISAs). In 2008, 12 laboratories measured all three markers, three laboratories measured one or two marker(s) by ELISAs and three laboratories measured the markers by Luminex.
Results
In 2004, the ELISA intercentre coefficients of variance (interCV) were 31%, 21% and 13% for Aβ 1–42, Tau and P-Tau, respectively. These were 37%, 16% and 15%, respectively, in 2008. When we restricted the analysis to the Innotest® (N = 13) for Aβ 1–42, lower interCV were calculated (22%). The centres that participated in both years (N = 9) showed interCVs of 21%, 15% and 9% and intra-centre coefficients (intraCV) of variance of 25%,18% and 7% in 2008.
Conclusions
The highest variability was found for Aβ 1–42. The variabilities for Tau and P-Tau were lower in both years. The centres that participated in both years showed a high intraCV comparable to their interCV, indicating that there is not only a high variation between but also within centres. Besides a uniform standardization of (pre)analytical procedures, the same assay should be used to decrease the inter/intracentre variation.
Introduction
A pronounced increase in the cerebrospinal fluid (CSF) levels of Tau and Tau phosphorylated at threonine 181 (P-Tau) proteins, and a significant decrease in the CSF level of amyloid-beta 1–42 (Aβ 1–42), has been found in Alzheimer disease (AD) patients compared with controls. 1–3 For this reason measurement of Aβ 1-42, Tau and P-Tau in CSF has gained wide recognition not only to discriminate AD patients from controls or patients with other types of dementia 4–6 but also to detect incipient AD in mild cognitive impairment (MCI) patients. 7,8
The number of laboratories that measure these parameters has increased substantially in the past few years. Different laboratories report different concentrations, resulting in differences in reference ranges and reference values. 5,9–14 Variations are partly due to differences in the enzyme-linked immunosorbent assay (ELISA) systems, methods and protocols used. Even if the same ELISA is used, inter-technician variations, differences in performing intra-centre validations and differences in the production of commercially available ELISAs over time can induce variability. 15,16 Finally, preanalytical factors such as differences in sampling, processing and storage of CSF can contribute to intra and intercentre variations. 17–19
External quality control measurements showing acceptable intra/intercentre variations are needed to compare intercentre biomarker results in multicentre studies. 20,21 However, to date, only one study has performed a quality control programme between 14 centres in Europe and reported intercentre variations ranging from 26% to 29% for Aβ 1–42, Tau and P-Tau. 9 In this study we report a quality survey performed in 20 laboratories all over the world involving two time points: one in 2004 and one in 2008.
Methods
Participants
Participants were requested to cooperate, and Aβ 1-42, Tau and P-Tau were measured in three different CSF-pool samples. They were again contacted before sending the samples. The packages were shipped on dry ice with a 24-h shipping service only after confirmation. The laboratories were asked to contact the Clinical Chemistry laboratory of the VUmc on arrival of the packages, to report the status of the arrived samples. If the samples were not frozen on arrival, new samples were sent.
Samples
In 2003, three different CSF-pool samples were provided by the Neurochemistry Laboratory of Möldal Hospital Sweden, to the Clinical Chemistry Laboratory of the VUmc. At least 150 aliquots of each sample were sent in dry ice and subsequently stored at −80°C. Sample 1 had high Tau and P-Tau levels, sample 2 had a low Aβ 1-42 level and sample 3 had a normal biomarker profile. These three CSF-pool samples were distributed to 13 laboratories in 2004 and the same samples were again distributed to 18 laboratories in 2008. Before shipping, the samples were selected, screened and labelled by two technicians at the laboratory of the VUmc. The samples were blinded for the participating laboratories and were sent worldwide on dry ice with the instruction to store the samples at −80°C.
CSF analysis
In 2004, 13 centres participated in the quality control programme. Aβ 1–42, Tau and P-Tau were measured by seven laboratories and six labs measured one or two parameter(s) by ELISA. All Tau and P-Tau levels were measured by the Innotest® (Innogenetics NV, Gent, Belgium). Aβ 1–42 was measured by eight centres with the Innotest®, 22–24 two labs used the Genetics Company® (Schlieren, Switzerland) test 25 and one lab ran an in-house assay. One laboratory showed the opposite pattern of the expected biomarker results, probably due to sample switching. These results were excluded from the analysis.
In 2008, 18 centres participated in the quality survey. Aβ 1–42, Tau and P-Tau were measured by 15 laboratories and three labs measured one or two biomarker(s) (see Table 1). To measure Aβ 1–42, 13 centres used the Innotest®, three laboratories used the Luminex (AlzBio®, Innogenetics NV, Gent, Belgium), one lab ran an in-house assay and one lab performed the Biosource® assay (Invitrogen, Camarillo, USA). 26 To measure Tau and P-Tau, 13 centres used the Innotest® and three labs used the Luminex (AlzBio®). 27
Results of the three biomarkers measured at two time points
Sample 1, high Tau and P-Tau profile; Sample 2, low Aβ 1–42 profile; Sample 3, normal biomarker profile; ELISA, enzyme-linked immunosorbent assay; interCV, intercentre coefficients of variance
Tau and P-Tau levels were all measured by the Innotest in both years. In 2004, nine centres measured Aβ 1–42 with the Innotest®, two labs with the Genetics Company® and one lab performed an in-house assay. In 2008, 13 centres measured Aβ 1–42 with the Innotest®, one lab with an in-house assay and one lab performed the Biosource® assay. Three laboratories measured all three markers by Luminex (AlzBio®)
Of the 11 centres that participated in both years, nine labs performed all measurements with the same assay. These nine centres used the Innotest® for Tau and P-Tau measurements. Aβ 1–42 was measured by eight centres with the Innotest® and one lab used an in-house test.
According to the data provided by the manufacturers of the Innotest®, the intra-assay variations are less than 6%, and the inter-assay variations are less than 10% for Aβ 1–42, Tau and P-Tau. For the Luminex (AlzBio®), the intra-assay variations are less than 4% and the inter-centre variations are less than 10%, for the three biomarkers. 27
Statistics
Data were analysed with the SPSS software package (version 15.0 for Windows SPSS, Chicago IL, USA). The Mann–Whitney U test was used to compare the CSF levels between 2004 and 2008. The intercentre coefficients of variation (interCV) were calculated according to the following formula: (SD)/(mean of the measurements) for each biomarker and each sample. The intracentre coefficients of variances (intraCV) were calculated according to the following formula: mean of (SD of the two measurements)/(mean of the two measurements). Subsequently, we calculated the mean CV of the three samples for each biomarker.
Results
Table 1 shows the median (min–max) values for the three biomarkers of the three samples measured in 2004 and 2008. There were no significant differences on comparing the ELISA results of 2004 and 2008, except for Tau levels of sample 3, which showed a decrease (P < 0.05).
Figure 1 shows the individual biomarker values of the three samples in both years. Luminex (AlzBio®) results are shown with squares and ELISA measurements are shown as dots. Besides large differences in absolute CSF levels between these two methods (Table 1), a lack of linearity was seen, 28 indicating that the differences cannot be attributed solely to standardization, rendering comparison of these two methods not useful.
The individual biomarker level for the three different CSF samples measured by the different centres in both years
The ELISA interCVs for Aβ 1-42, Tau and P-Tau were relatively high, with comparable variability in both years (Table 1). When only the Innotest® was used to measure Aβ 1–42 (N = 13), the mean interCV of 2008 (22%) was substantially lower. (Note that all centres used Innotest to measure Tau and P-Tau.) Luminex® mean interCVs were comparable to Innotest interCVs for all three biomarkers.
The centres that participated in both years (N = 9) showed an ELISA interCV of 30%, 21% and 13% in 2004, and 21%, 15% and 9% in 2008. IntraCVs were 25%, 18% and 7% for Aβ 1–42, Tau and P-Tau (individual values are shown in Figure 1 by connecting lines).
Discussion
The main finding of this study is that the intercentre variability in three widely used CSF biomarkers for AD was high. The highest intercentre variability was found for Aβ 1–42. The variabilities for Tau and P-Tau were lower. Further, the intracentre variability for the three biomarkers, calculated for the centres that participated in both years, was comparable with the intercentre variability, indicating that there is not only a considerable variation between but also within centres. Considering the fact that for CSF markers external quality assessment schemes (EQAS) are still not available, this result is encouraging. 29
There are several potential reasons for the high variation of the Aβ 1–42 measurement found in 2004 and 2008. Different antibodies, incorporated into the different ELISA assays, catch/detect different kinds and different amounts for Aβ 1–42, resulting in different calculated values of this peptide in CSF. In this study several assays (Biosource®, Genetics Company®, Innotest®, in-house assays) have been used, resulting in different values and thus a higher variation. When we restricted the analysis only to the Innotest® or to the Luminex® (AlzBio3®) results, lower variations were observed. However, the intercentre variability for all three biomarkers still remained relatively high in comparison with the data provided by the Innotest® manufacturer, indicating that there are other factors that may have caused the variability. Several publications have shown that preanalytical factors such as sampling in different tubes, contamination with blood, processing-time after withdrawal, time of lumbar puncture and different storage conditions influence the measured concentration of amyloid in CSF. 10,17–19,30,31 However, these conditions were unlikely to have caused the variability, since all samples were processed/collected/aliquoted in the same manner and all aliquots were directly stored at −80°C in polypropylene tubes. Variability can also be induced by analytical factors such as differences in (room) temperature while running the assay, differences in the incubation time, the use of manual or mechanical pipetting systems, intertechnician variations in adherence to the assay procedure and differences in detecting machinery. However, few effects in the analytical results due to these factors have been reported in the literature. 10,17 In addition, in this study, the measurements were performed in experienced laboratories, making it less probable that these factors were responsible for much of the variation.
Finally, variability can also be induced by assay lot-to-lot differences. This has been shown by our group, which ran the same assay using the same CSF at two different time points. 15,32 These studies have shown a high degree of variability in the results of measurements of these three biomarkers, indicating that there are probably differences in the assays prepared by the same manufacturer over time (differences in the production of antibodies, in the standard solutions and in the plates used). Unfortunately, we do not know which batch number the different labs used in this study, but this may be an important cause of variability between and perhaps, even more importantly, also within centres.
Large differences in the absolute biomarker levels between centres have also been reported in a meta-analysis. 14 In addition, the high ELISA intercentre variability found for Aβ 1–42 is in line with the results of Lewczuk et al. 9 However, the intercentre variations for Tau and P-Tau are considerably lower in our study than in this former multicentre comparison, showing that lower variability can be attained, thus encouraging further studies.
In summary, the high variability found, especially for Aβ 1–42 (>22%), has clinical implications and lower variability (<10%) is required to reliably distinguish a difference in groups. For commonly used immunoassays, EQAS, with over 100 participants, intercentre variabilities of 9–15% (TSH and Prolactin) have been reported (data reported by the Dutch National External Quality Assessment Scheme for Ligand-Assays of Hormones, Tumor markers and Vitamins). From this perspective, together with the information from the assay manufacturers and the within-laboratory variations reported in the literature, 15,32 lower intercentre variabilities should be obtained for Aβ 1-42, Tau and P-Tau. This is required to make multicentre biomarker comparison possible and eventually to obtain an intercentre reference range for AD.
DECLARATIONS