Statistics from Altmetric.com
Pisetsky et al reported on assay variation in the detection of antinuclear antibodies (ANA) in sera of patients with established systemic lupus erythematosus (SLE).1 The authors determined ANA in sera from 103 patients with established SLE using three different validated and widely used immunofluorescence assays (IFA) (from ImmunoConcepts, Inova Diagnostics and Bio-Rad), an ELISA and a bead-based multiplex assay (both from Bio-Rad). With IFA, the frequency of ANA negativity varied from 4.9% to 22.3%. Part of the samples (11.7% and 13.6%) were negative with ELISA and multiplex assay, respectively. This study showed differences among IFA ANA kits which might have implications for eligibility for clinical trials.1
In a correspondence to the paper of Pisetsky et al.1 Dr Mahler discussed some relevant points related to ANA detection,2 including the substantial interobserver variability inherent to IFA. Accordingly, Dr Mahler recommended to use automated interpretation systems to reduce variability and subjectivity.2
We evaluated variation in ANA detection by automated IFA systems. Three samples (sample 1: homogeneous 1:320; sample 2: fine speckled 1:80; sample 3: negative) were distributed to 31 Belgian laboratories that use automated IFA systems and 27 laboratories participated (NOVA View n=12 (Inova Diagnostics, San Diego, USA); EUROPattern n=6 (Euroimmun, Lübeck, Germany); G-Sight n=7 (Menarini, Firenze, Italy); Image Navigator n=2 (ImmunoConcepts, Sacramento, California, USA)). Each sample was determined at least four times in different runs (the majority was determined at least eight times) according to the instructions of the manufacturer. The fluorescence intensity units (reported as light intensity units (LIU) or probability index) are shown in figure 1, except for EUROPattern that does not report fluorescence intensity. The results not only show variation in the way fluorescence intensity is reported by different companies but also (1) variation in between-run imprecision between automated IFA systems from the same manufacturer and (2) variation in fluorescence intensity results between instruments from the same manufacturer. For instance, for NOVA View, the coefficient of variation for sample 1 varied between 8.6% and 39.7% and there were statistically significant differences (after correction for multiple comparisons) in LIU values between laboratories (figure 1 and its legend). Also for G-Sight, significant differences between laboratories were found. For sample 2, six laboratories using NOVA View and three laboratories using G-Sight found the sample to be negative in 10%–80% of the between-run determinations for NOVA View and in 12.5%–62.5% of the between run determinations for G-Sight. The other laboratories found the sample to be positive for all determinations, with differences in fluorescence intensities between laboratories. For sample 3 (negative sample), three laboratories using NOVA View and four laboratories using G-Sight found positive results in 10%–62.5% of the determinations for NOVA View and in 10%–100% of the determinations for G-Sight. Some of these results were due to artefacts as expert visual review scored 60%–80% of the false positive determinations for NOVA View negative.
For a selection of patterns, NOVA View allows to estimate the end titre from the analysis of the 1:80 dilution (see legend of figure 1). EUROPattern (performed by six laboratories) does not provide a measure of fluorescence intensities. It scored sample 1 positive for all determinations (n=61) (with the estimated end titre varying between 1:80 and 1:1280) and sample 2 for 59% of the determinations (n=51) (with the estimated end titre varying between <1:80 and 1:160). Sample 3 was negative in 92% of the determinations (n=60).
Pattern assignment by automated IFA was correct in 76%, 95% and 100% of the determinations by NOVA View, EUROPattern and G-Sight for sample 1. For sample 2, the values were, respectively 32%, 100% and 100%. This indicates that pattern assignment by automated systems should be verified by an experienced technician or immunologist.
Sample 1 had been used to evaluate interinstrument variability for NOVA View in a previous study3 performed 1 year before the current study. This gave us the possibility to evaluate constancy of the results over a period of 7–19 months for 11 of the 12 NOVA View users. The results are represented in figure 2 and show statistically significant differences between LIU values obtained at two different time points for 6/11 laboratories. Of note, for five laboratories, no statistical significant differences between the two determinations were observed, which illustrates the potential of automated IFA analysis to provide reproducible results over a longer time period.
Taken together, although we could demonstrate reproducible ANA results for some laboratories, our results indicate variation in ANA detection by automated IFA systems. We not only found variation between automated IFA analysis using instruments from different manufacturers but also between instruments from the same manufacturer. Efforts should be undertaken to harmonise automated IFA analysis. This could include the use of standards, calibration of the instruments and monitoring of the quality of the slides and reagents.
We thank Sofie Arens (Klinisch laboratorium Declerck Ardooie), Jan Beckers (AZ Sint-Blasius Dendermonde), Mario Berth (AML Antwerpen), Juul Boes (AZ Turnhout), XB (UZ Leuven), Francis Corazza and Carole Nagant (CHU Brugmann Brussel), Bjorn Ghesquière and Ignace Van Hecke (CRI Zwijnaarde), Sylvie Goletti (UCL Louvain), An Joosten and Patricia Meirlaen (UZ Brussel), Julia Kovaleva (CMA Hasselt), Laurence Lutteri (CHU Liège), Laurence Miller and Patricia Evrard (CHU UCL Namur), An Nijs (AZ Groeninge Kortrijk), Mounzer Reda (Hôpital de Jolimont Haine-Saint-Paul), SS (GZA Antwerpen), Rita Schroder (Vivalia Arlon), Chistine Tilmant (C.H.I.R.E.C. Braine l’Alleud), Tanja Troch (ASZ Aalst), Wim Uyttenbroeck (ZNA Jan Palfijn Antwerpen), LVH (OLVZ Aalst), Hilde Vanhouteghem (AZ Maria Middelares Gent), Annemie Van Ruymbeke (AZ Alma Sijsele), Lut Vanneste (Labo Bruyland Kortrijk), Martine Vercammen (AZ Sint-Jan Brugge), Ann Verdonck (Anacura Labo Nuytinck Evergem) and Sara Vijgen (Jessa ziekenhuis Hasselt) for participating in the study. We thank Heide De Baere and Nathalie Ruisseau (Menarini Benelux), Tom Gheskiere (Biognost cvba), Chris Raskin (Biomedical Diagnostics N.V./S.A Benelux) and Nathalie Vandeputte (Werfen N.V./S.A Benelux) for their practical support.
Funding The authors have not declared a specific grant for this research from any funding agency in the public, commercial or not-for-profit sectors.
Competing interests XB has been a consultant for Inova Diagnostics.
Patient consent Not required.
Ethics approval Local ethical committee OLVZ: amendment for study 2015/100 (B126201525864) received on 07/03/2017.
Provenance and peer review Not commissioned; internally peer reviewed.
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.