Anti-dsDNA Testing Specificity for Systemic Lupus Erythematosus: A Systematic Review

Background: Autoantibody speciﬁcity in autoimmune diseases is variable due to each patient’s individual spectrum of autoantibodies and the inherent differences between detection methods and tests. Since false-positive results have downstream consequences, we conducted a comprehensive assessment of anti-double stranded DNA (anti-dsDNA) speciﬁcity from published studies of systemic lupus erythematosus (SLE). Methods: A systematic review (MEDLINE, Embase, Cochrane Central Register of Controlled Trials, and Database of Abstracts of Reviews of Effects) identiﬁed cross-sectional or case-control studies published January 2004 to August 2019, reporting anti-dsDNA test accuracy data in SLE. Study quality was assessed using Quality Assessment Tool for Diagnostic Accuracy Studies, version 2. A meta-analysis was conducted to estimate speciﬁcity by test method or named test where feasible. Results: Thirty studies were included covering 43 different tests. The Crithidia luciliae indirect immunoﬂuorescence test (CLIFT) and ﬂuorescence enzyme immunoassay methods are likely to be (cid:2) 90% speciﬁc (Euroimmun 97.8% (95% CI 96.2%–98.7%) 4 studies; EliA 94.7% (95% CI 91.7%–96.7%), 6 studies; CLIFT 98.7% (95% CI 96.7%– 99.5%), 8 studies/7 tests]. For other test methods, speciﬁcity was not fully demonstrated to be (cid:2) 90% and/or the control group included healthy patients possibly overestimating speciﬁcity. More studies are required for NOVA Lite [96.0% (95% CI 87.2%–98.9%), 5 studies], chemiluminescence immunoassays [92.3% (95% CI 83.6%–96.6%), 6 studies/4 tests], multiplex immunoassays [89.3% (95% CI 86.1%–91.8%), 4 studies/2 tests], and Farr ﬂuorescent immunoassays (no estimate, 2 studies). Speciﬁcity data reported for Farr radioimmunoassays [93.8% (95% CI 85.4–97.5%), 11 studies, 9 tests] and enzyme-linked immunosorbent assays [93.4% (95% CI 89.9%–95.7%), 15 stud-ies/16 tests] lacked consistency. Conclusion: Anti-dsDNA testing shows considerable variation in test speciﬁcity, with potential impact on the management of SLE patients. This review may help laboratory specialists and clinicians choose and interpret the appropriate anti-dsDNA test for their setting. Speciﬁc anti-double stranded DNA (anti-dsDNA) testing is crucial for accurate classiﬁcation and diagnosis of systemic lupus erythematosus, a potentially life-threatening systemic autoimmune disease and might help in disease activity assessment. Currently, different anti-dsDNA test options are available with predicted quality differences in test performance, affecting test speciﬁcity. Our comprehensive assessment of available anti-dsDNA tests, based on the systematic review of relevant accuracy studies, indeed conﬁrms mark-edly variable test performance, potentially resulting in > 10% false-positive rates. The overview presented can guide laboratory specialists in the optimization of anti-dsDNA test strategy, improving patient management and disease outcomes. erythematosus; EULAR, League Against SLICC, Systemic Lupus dou-ble-stranded DNA; Farr-RIA, Farr radioimmunoassay; Farr-FIA, Farr CLIFT, Crithidia im- munofluorescence test; FEIA, fluorescence enzyme CLIA, immunoassay; immunoassay; summary characteristic;


INTRODUCTION
Systemic lupus erythematosus (SLE) is a heterogeneous disease (1). With positivity for antinuclear antibodies (ANA) as a major entry criterion, the 2019 European League Against Rheumatism (EULAR)/ American College of Rheumatology (ACR) SLE classification criteria stipulate higher specificity compared a to the previous Systemic Lupus International Collaborating Clinics (SLICC) criteria developed in 2012 (2,3). Although classification criteria are primarily intended for patient selection in clinical trials (4), ANA testing is also a routine tool in the general diagnostic workup for SLE due to its high sensitivity (2,5). As double-stranded DNA (dsDNA) antibodies are among the most commonly detected ANAs in SLE, reported to correlate also with disease activity and especially renal involvement (6)(7)(8)(9), a positive ANA test is routinely followed up with testing for these antibodies (2,10).
Many test methods are available and anti-dsDNA tests, by design, detect different subpopulations of anti-dsDNA with distinct fine specificities (16,20,(22)(23)(24). One of the original methods is the Farr radioimmunoassay (Farr-RIA), first described in 1969 and sometimes still referred to as the gold standard (25). The technique relies on immunoprecipitation with radiolabeled dsDNA under high salt conditions to select for highly avid anti-dsDNA antibodies (26). To circumvent safety issues related to handling of radioactive reagents, Farr fluorescent immunoassays (Farr-FIA) use fluorescence as a readout (27). Another established, fluorescence-based method used for almost 5 decades is the Crithidia luciliae indirect immunofluorescence test (CLIFT), which makes use of the compact arrangement of dsDNA in the mitochondrion of the Crithidia cell, the so-called kinetoplast (28,29). This method is laborious and requires skilled technicians, and its routine use in laboratories varies widely (30). Enzyme immunoassays have helped to overcome limitations such as userdependent variability, processing time, and safety concerns and have led to partial or even full automation. ELISAs are typically performed in a microtiter well with the antigen of interest immobilized to a solid surface, with quantification of bound antibodies through a colorimetric signal generated by an enzyme coupled to a detection

IMPACT STATEMENT
Specific anti-double stranded DNA (anti-dsDNA) testing is crucial for accurate classification and diagnosis of systemic lupus erythematosus, a potentially life-threatening systemic autoimmune disease and might help in disease activity assessment. Currently, different anti-dsDNA test options are available with predicted quality differences in test performance, affecting test specificity. Our comprehensive assessment of available anti-dsDNA tests, based on the systematic review of relevant accuracy studies, indeed confirms markedly variable test performance, potentially resulting in > 10% false-positive rates. The overview presented can guide laboratory specialists in the optimization of anti-dsDNA test strategy, improving patient management and disease outcomes. antibody. Fluorescence enzyme immunoassays (FEIA) and chemiluminescence immunoassays (CLIA) are variations of this principle using fluorescence or luminescence as a readout resulting in a higher analytical measurement range and a reduced need for sample dilutions. CLIA typically use paramagnetic beads as the solid phase for antigen immobilization. Multiplex immunoassays (MIA) combine the bead format with multiple fluorescent dyes to enable the simultaneous detection of antibodies to multiple autoantigens. To limit the binding of antibodies that may be of little or no clinical relevance, manufacturers apply different approaches (31). Despite this, there is still a risk of false positives with any anti-dsDNA test, potentially leading to further investigations to rule out SLE.
When developing the 2019 classification criteria for SLE, the EULAR/ACR noted that some anti-dsDNA tests used in practice have relatively low specificity (2). Hence the EULAR/ACR criteria set out a benchmark for the specificity of anti-dsDNA testing, that is "an immunoassay with demonstrated ! 90% specificity for SLE against relevant disease controls" (2). This statement has 2 implications: (i) "with demonstrated ! 90% specificity" implies there is a need to account for all the evidence available for each test, and (ii) "against relevant disease controls" implies that the specificity should be measured in a range of controls that is representative of the population that would receive the test in practice.
On this basis, we set out to conduct a systematic literature review that would identify all published studies evaluating the diagnostic accuracy of anti-dsDNA tests for SLE. This provides a comprehensive evidence base upon which the discussion in this paper is based. We provide an overview of the specificity data reported in these studies and a formal assessment of the quality of the studies, including an appraisal of the control group in which specificity is measured. In addition, we conducted a meta-analysis to pool data across studies to assess whether anti-dsDNA tests are likely to meet a benchmark ! 90% specificity in practice.

Screening of Studies for Inclusion in the Review
All citations retrieved from the electronic search were screened against the prespecified study inclusion criteria (see Supplemental Table 1 on the online Supplemental Data for details). Eligible studies were diagnostic test accuracy studies with a crosssectional or case-control design and at least 10 patients enrolled. The study populations needed to include an SLE cohort (cases) and a non-SLE group (controls) in which specificity was measured. All patients had a definitive classification of SLE (for The initial screening of the studies was based on the citation title and abstract, with a second screening using full-text papers to confirm their eligibility for inclusion in the review. The literature search citation flow is reported as per the Preferred Reporting Items for Systematic reviews and Metaanalyses of Diagnostic Test Accuracy studies statement (34) and is provided in Supplemental Fig. 2. A summary of the included studies and test information is provided in Supplemental Table 4.

Study Data and Summary Estimates of Specificity
Each study needed to report the total number of true positives, true negatives, false positives, and false negatives for each test. Specificity data from each study are grouped by test method and presented in forest plots (Figs. 1 and 2) with a 95% CI calculated using the exact (Clopper-Pearson) method (35). If studies reported data separately for healthy and diseased controls, the calculation of specificity was based on results from disease controls only (see section Relevance of Disease Controls).
In addition, hierarchical summary ROC (HSROC) curve plots of sensitivity vs specificity (expressed as 1 À specificity) were generated after fitting the test count data (36,37). The plots show the individual study data and a summary (pooled) estimate taking into account all studies as well as the HSROC curve, a 95% confidence region around the summary estimate (based on the study observations), and a 95% prediction region [to allow for potential unobserved heterogeneity: if a new study was conducted we would expect the "true" sensitivity and specificity to lie within the prediction region with a 95% confidence level (32,37,38)]. The summary HSROC graphs are presented in Supplemental Figs. 6 to 8. All analyses were conducted in STATA MP v16.1. Note that a minimum of 4 studies are required for the HSROC plot.

Relevance of Disease Controls
In this paper, the assessment of whether studies measured specificity in a relevant disease control group was guided by the derivation and validation cohorts used to develop the 2019 EULAR/ACR classification criteria (see Supplemental Table 3) (2). International SLE experts contributed control samples from patients evaluated at their centers, with a range of medical conditions mimicking SLE including other connective tissue diseases (Sjögren syndrome, systemic sclerosis, inflammatory myopathies, undifferentiated connective tissue disease, or mixed connective tissue disease) and other rheumatic diseases (rheumatoid arthritis, psoriatic arthritis, spondyloarthritis, and osteoarthritis). Healthy controls (blood donors or volunteers with no relevant medical condition) do not represent patients who would receive anti-dsDNA testing in practice, and their inclusion in the control group could lead to an overestimation of anti-dsDNA test specificity (39).

Quality Assessment
The study quality was assessed using the Quality Assessment Tool for Diagnostic Accuracy Studies, version 2 (QUADAS-2) checklist (40) to assess the potential for bias in patient selection, attrition, flow, and timing of the tests and conduct and interpretation of the index tests and reference standard. The QUADAS-2 assessment for each study included in the review is given in   Table 5, and Fig. 3 provides an overview of the quality of the studies.  Forest plot of anti-dsDNA test specificity by using the ELISA, Farr-RIA, or Farr-FIA method. Red line 90% specificity, grey box 5 study includes healthy controls (specificity may be overestimated in this study); purple box 5 study includes only healthy controls (specificity may be overestimated in this study). Excluded 5 95% CI could not be estimated using the exact method as FP 5 0. TN, true negatives; FP, false positives.

Studies Included in the Review
CLIFT (42,47,55,68)  CLIA. Three out of the six studies with data for CLIA tests reported a specificity point estimate less than 90% (range 89%-98% for QUANTA Flash; 71%-96% for other CLIAs). Based on the data available for the four different CLIA tests at the time of this review, the summary estimate for specificity is 92.3% (95% CI 83.6%, 96.6%), (see Supplemental Fig. 7, A). It is noted that 15.1% of control subjects were healthy controls, and for two studies the disease controls may not be representative of controls in practice (in comparison to the EULAR/ACR 2019 derivation and validation set).
CLIFT (Euroimmun at cutoff of 1:10). All four studies reported specificity above 90% although two studies included healthy controls in the specificity estimates: these studies reported higher specificity compared to the two studies that included  • Three out of the 6 studies reporting specificity for CLIA report a point estimate less than 90%.
• Specificity of the QUANTA Flash CLIA test is not currently demonstrated to be ! 90% nor demonstrated against relevant disease controls; more studies are required to conduct a meta-analysis for this named test and 2 out of 3 studies included healthy controls.
• Specificity of tests using the CLIA method in general is not demonstrated to be ! 90%: HSROC summary estimate incorporating 6 studies, lower 95% CI is less than 90%.
• Not demonstrated against relevant disease controls: 11% of control subjects were healthy controls and for 2 studies the disease controls may not be representative of controls in practice (does not include the broad range of patients with an alternative diagnosis as seen in the EULAR/ACR 2019 derivation and validation set).
• Caveats: all of the studies were in secondary/tertiary care settings and most were of a case-control design CLIFT • 22% of control subjects were healthy controls, and for 1 study the disease controls may not be representative of controls in practice (does not include the broad range of patients with an alternative diagnosis as seen in the EULAR/ACR 2019 derivation and validation set).
• Specificity of the Euroimmun CLIFT is likely to be ! 90% based on HSROC summary estimate and 95% CI but is not fully demonstrated against relevant disease controls. Specificity may be lower in practice than that observed in the studies.
• Caveats: all of the studies were in secondary/tertiary care settings, and most of the studies were of a case-control design. CLIFT • None of the 3 studies reporting specificity for CLIFT ImmunoConcepts † report a point estimate less than 90%.
• Specificity of the Immuno Concepts CLIFT test is likely to be ! 90% but more studies are required to conduct a meta-analysis for this named test.
• Specificity for other CLIFTs likely to be ! 90% based on the HSROC summary estimate incorporating 8studies and 95% CI, but most tests only have data from 1 study data, and specificity is not fully demonstrated against relevant disease controls: 17% of control subjects were healthy controls, and for 1 study, the disease controls may not be representative of controls in practice.
• Caveats: All of the studies were in secondary/tertiary care settings, and most of the studies were of a case-control design. ELISA  Fig. 6, B). It was noted that 11% of control subjects were healthy controls. Studies including healthy controls reported higher specificity compared to the studies that included disease controls only ( Fig. 1 and HSROC in Supplemental Fig. 6, B). For two studies the disease controls may not be representative of controls in practice (in comparison to the EULAR/ ACR 2019 derivation and validation set).

CLIFT (other tests).
Eight studies reported data for seven other CLIFT tests, all of which had specificity above 90%. Specificity of the Immuno Concepts CLIFT tests (Fluorescent nDNA and one unnamed) is likely to be ! 90% (reported range 97%-100% in disease controls with similar range of patients to EULAR/ACR 2019 derivation and validation set): more studies are required to conduct a metaanalysis for this test. For the other CLIFT tests, two studies included healthy controls in the specificity estimates. Studies including healthy controls tended to report higher specificity compared to the studies that included disease controls only ( Fig. 1 and HSROC in Supplemental Fig. 7, B). In one study the disease controls may not be representative of controls in practice (in comparison to the EULAR/ACR 2019 derivation and validation set). Based on the data from eight studies covering seven different CLIFT tests at a cutoff 1:10, the summary estimate for specificity is 98.7% (95% CI 96.7%-99.5%) (see Supplemental Fig. 7, B).
ELISA. Specificity of the ELISA tests is not demonstrated to be ! 90% based on the lack of consistent data for this test method. Most of the ELISA tests only had data from one study, and there is a wide variation in test characteristics making comparisons difficult. Reported specificity ranged from 71% to 100%. Four studies (covering six ELISA tests) reported a point estimate less than 90% for specificity. Based on the data from 15 studies covering 16 different ELISA tests at various cutoffs, the summary estimate for specificity is 93.4% (95% CI 89.9%-95.7%) (Supplemental Fig. 8, A). Specificity was not always measured against relevant disease controls: 20.8% of control subjects were healthy controls, and for two studies, the disease controls may not be representative of controls in practice. Five studies reported data for an Inova ELISA (specificity range 90%-96%); however, these data were not compared in a separate meta-analysis given the mixed test characteristics.
Farr-FIA. Data were available from two studies. More studies are required to conduct a metaanalysis for Farr-FIA. Specificity was reported to be 95% to 100%; however, this could be an overestimate given that the majority of control patients were healthy controls (72.4% healthy blood donors).
Farr-RIA. Three out of the 11 studies reporting specificity for Farr-RIA report a point estimate less than 90% with one study reporting a specificity as low as 53%. Specificity is not demonstrated to be ! 90% based on the HSROC summary estimate incorporating 11 studies [93.8% (95% CI 85.4%-97.5%)] (see Supplemental Fig. 8, B). This is based on the 95% CI (also the large 95% prediction region) and the lack of consistency in test characteristics. Most tests only have data from one study, and specificity is not fully demonstrated against relevant disease controls with four studies including 67% to 100% healthy controls.
FEIA (EliA dsDNA at cutoff of 15 IU/mL). All six studies reported specificity above 90%, and healthy controls were excluded from the specificity estimates. MIA. The three studies reporting specificity for MIA Bioplex reported a point estimate less than 90% (range 86.7%-89.8%). All three studies included healthy controls. More studies are required to conduct a meta-analysis for MIA Bioplex. One additional study reported data for FIDIS (specificity 91.9%). Combining the data from these four studies of MIA anti-dsDNA tests, the summary estimate for specificity is 89.3% (95% CI 86.1%, 91.8%) (see Supplemental Fig. 7, C). More studies are required to demonstrate the specificity of MIA especially studies in relevant disease controls. Selection of the patients. It was noted that most of the studies (90%) were a case-control design, and studies with this design are downgraded in the QUADAS-2 assessment. Whilst this may lead to selection bias, some case-control studies enrolled consecutive patients for each group or avoided overly restrictive exclusions.

Summary of Study Quality Assessment
For this review we assessed whether the disease control patients enrolled in the study represented a relevant control group for measuring test specificity. The control group was deemed to be representative to the review question in 70% of studies. Nine studies included healthy controls or did not include a range of relevant disease controls (see Supplemental Table 4). It was noted that patients were selected from patients presenting for anti-dsDNA testing in secondary/tertiary care settings in most studies. The majority of studies were conducted in European centers (21 studies), with 3 studies in China, 4 studies in other Asian countries, and 3 United States/ Canada studies.
Index test. There were 2 areas where the conduct or interpretation of the index test results may bias the results. One of the QUADAS-2 signaling questions is whether the index test results are interpreted without knowledge of the reference standard (not known if patient is in the SLE cohort or control group). Case-control studies have been assessed to have a high risk of bias for this domain unless the publication mentioned that index test results were interpreted without knowledge of the reference standard. The second signaling question is whether the test threshold was prespecified. For 1 study, the test threshold was based on maximizing the ROC area under the curve; otherwise, the test cutoff was set before the test was conducted. We note that the results from anti-dsDNA tests are (semi-)quantitative, and we did not note any particular issues with  . Although some studies have been marked down in the QUADAS-2 assessment, it is unlikely that interpretation bias is an issue. Note that 1 test result was excluded from the analysis of CLIFT tests. In this study the specificity of CLIFT NOVA Lite was reported to be 44% (69), which is an outlier compared with other CLIFT anti-dsDNA test data (42, 44, 47, 50-52, 55, 56, 60, 61, 66-69).
Reference standard. The included studies used ACR 1982 (33), ACR 1997 (13), SLICC 2012 (3), or a combination of these classification criteria to define the SLE group. We therefore deemed all studies to have used an acceptable reference standard to determine the target condition. Similarly, the control group of patients had a definitive diagnosis other than SLE. One of the signaling questions relates to the incorporation of the index test into the reference standard. This potential form of confirmation bias was noted in 60% of studies where it was assumed that an anti-dsDNA test was part of the diagnostic workup. For other studies the index test was a different test to that used for diagnosis, or the publication indicated that the reference standard was independent of the index testing.
Flow and timing. For the final domain the signaling questions relate to the time interval between the index test (or more accurately the time when serum sample was taken) and reference standard, differential or partial diagnosis, and handling of missing results (unexplained exclusion of patients or test results, or uninterpretable or equivocal test results). All studies were judged to have a low risk of bias for this domain.
To summarize, the key area of potential bias arises from the selection of the patients, in particular the use of case-control designs and range of disease controls included in the studies.

DISCUSSION
This review includes 30 studies covering 43 different anti-dsDNA tests, which we have grouped by method for comparison purposes. It should be noted that systematic reviews are designed to include all the published evidence available for each test that meets the inclusion criteria. This evidence base, and the resultant meta-analysis of the data confirms that there is a wide variation in anti-dsDNA test performance and effectiveness for classification and diagnosis of SLE. Some studies report false-positive rates above 10% (out of 30 studies: 2 CLIA, 2 NOVA Lite CLIFT, 4 ELISA, 3 Far-RIA, 3 MIA). For some methods more data are needed (such as the newer methods CLIA, MIA, and Farr-FIA). The current data indicates that FEIA and some CLIFT tests are likely to have a specificity ! 90%: for these named tests the 95% CI from the meta-analysis was above 90%, and the disease control group was judged to be sufficiently relevant. It was noted that for all tests the 95% prediction region shown in the HSROC plots included values where specificity < 90%. If a new study was conducted, we would expect the "true" specificity to lie somewhere within the prediction region with a 95% confidence level. Despite the number of studies reporting test performance for Farr-RIA and ELISA, there is a wide range of test characteristics and a lack of consistency the data reported, such that we are unable to determine whether these tests are likely to have a specificity ! 90%. The Farr assay, first described for the detection of anti-dsDNA in 1969, is sometimes still regarded as the gold standard because of its high specificity for SLE. In addition, the Farr assay is the method that is included in the widely used composite index SLE Disease Activity Index 2000 (71). Our review of the data for Farr assays does not support the assay's status as a reference method for anti-dsDNA testing. It should be noted that there were too few data to allow estimates of individual tests for CLIA, MIA, Farr-RIA, and ELISA. The summary estimates in the paper are grouped according to analytical principle of the dsDNA method, although specificities clearly vary across assay manufacturer.
Eight studies in our review include between 20% and 100% of healthy subjects in their control groups, which can lead to an underestimation of the false-positive rate (39). Furthermore, it is important that diagnostic test accuracy studies include a relevant set of disease controls. This is particularly apt for antibodyspecificity testing in the context of autoimmune disease since the expression of anti-dsDNA comprises a mixture of antibodies and fine specificities (6,72). As part of our review, we contrasted the controls groups included in the studies against the validation set used for the EULAR/ACR 2019 classification criteria since this cohort included patients evaluated in practice, with a wide range of medical conditions mimicking SLE. More studies with a cross-sectional design would ensure that a broad range of relevant controls will be represented.
Anti-dsDNA tests of differing design may lead to the differential classification of patients, as is the case for other autoantibodies (e.g., rheumatoid factor and anti-CCP antibodies) (72)(73)(74). Furthermore, false-positive results have direct and indirect downstream consequences (75) and can result in a delay in obtaining the correct diagnosis and potentially inappropriate or no treatment. While the differences in the tests could hamper efforts toward test harmonization in autoimmune diagnostics (72), this is also an opportunity to adopt complimentary diagnostic approaches depending on the setting, such as a combination of a sensitive and a highly specific test (6).
The 95% CI estimates provided indicate the extent of variation in the specificity estimates reported in the studies. These are from a mixed-effect meta-analysis model that allows for variation within studies and between studies. The studies included the use of a single measurement with each test and for each patient. A different model would be needed to analyze the variance resulting from repeated testing of the same sera, although individual patient data are unlikely to be available to make this feasible. We noted that most of the studies included in our review were conducted in patients from secondary/tertiary care settings and mostly European countries. This may have an impact on the generalization of the results, for example, to the primary care setting where there is a low pretest probability of SLE and a need for a highly specific test. If there is a high suspicion of SLE, a more sensitive test may be required to support diagnosis. Our review was not designed to assess the sensitivity, and further work is required to assess sensitivity of anti-dsDNA testing to support the diagnosis of SLE, and separately anti-dsDNA testing to monitor disease activity.
In conclusion, this comprehensive assessment of anti-dsDNA test specificity has allowed us to estimate the likely variation in false-positive rates across tests. Some studies report false-positive rates > 10% for some of the tests, and since specificity is not always measured in an appropriate control group, this may be an underestimate. Given the complexity of antibody-specificity testing in autoimmune diseases, the evidence presented can help lab specialists identify the most appropriate anti-dsDNA tests for their setting. The review also identifies gaps in the evidence where more studies are required.

SUPPLEMENTAL MATERIAL
Supplemental material is available at The Journal of Applied Laboratory Medicine online. Author Contributions: All authors confirmed they have contributed to the intellectual content of this paper and have met the following 4 requirements: (a) significant contributions to the conception and design, acquisition of data, or analysis and interpretation of data; (b) drafting or revising the article for intellectual content; (c) final approval of the published article; and (d) agreement to be accountable for all aspects of the article thus ensuring that questions related to the accuracy or integrity of any part of the article are appropriately investigated and resolved.
Authors' Disclosures or Potential Conflicts of Interest: Upon manuscript submission, all authors completed the author disclosure form. Disclosures and/or potential conflicts of interest: Employment or Leadership: A. Voreck, Thermo Fisher Scientific, Phadia AB, Uppsala, Sweden; R. Aksouh, Thermo Fisher Scientific, Phadia AB, Uppsala, Sweden. Consultant or Advisory Role: None declared. Stock Ownership: R. Aksouh, Thermo Fisher Scientific, Phadia AB, Uppsala, Sweden. Honoraria: None declared. Research Funding: M.W.J. Schreurs reports grants from Thermo Fisher Scientific, during the conduct of the study; grants from Thermo Fisher Scientific, outside the submitted work. Expert Testimony: None declared. Patents: None declared. Other Remuneration: M.E. Orme reports personal fees from Thermo Fisher Scientific, during the conduct of the study; personal fees from Thermo Fisher Scientific, outside the submitted work.