A total of 964 participants consisting of 564 patients and 400 healthy donors were studied from January 2004 to June 2007. Of this total, 207 patients had SLE according to the updated and revised classification criteria of the American College of Rheumatology (ACR) 1920. Demographic data and a detailed characterization of the SLE patients are shown in Table S1 in Additional file 1.

The non-SLE cohort included 162 individuals with rheumatoid arthritis (RA) 21, 88 patients with Sjögren's syndrome (SS) who fulfilled the revised European classification criteria 22, 81 patients with SSc according to the ACR criteria of 1980 23 and 26 patients with myositis 24.

All patients were recruited from the Department of Rheumatology and Clinical Immunology, Charité University Hospital, Berlin, Germany. The Ethics Committee of the Medical Faculty of the Charité University Hospital approved the study, and written informed consent was obtained from all participants.

Sera from healthy donors were recruited in cooperation with the University of Lübeck and were investigated using the anti-dsDNA ELISA, antinucleosome ELISA and the Anti-dsDNA-NcX ELISA. The female:male ratio of healthy donors was 13:87, and their mean age was 35.13 years (age range, 18 to 65 years). Written informed consent was obtained from all healthy participants.

The Anti-dsDNA-NcX ELISA microtiter plates (Nunc, Roskilde, Denmark) were coated at 4°C first with a 0.1 μg/ml concentration of an ultrapure nucleosome preparation from calf thymus (free of Scl-70, histone H1 and other nonhistone components) 18 in sodium carbonate buffer for 3 hours, followed by a 1.5 μg/ml concentration of highly purified, native dsDNA isolated from calf thymus in sodium carbonate buffer overnight. After being washed with 0.05% phosphate-buffered saline (PBS)-Tween 20 (vol/vol) and blocked for 2 hours with 0.1% PBS (wt/vol) casein, sera diluted 1:200 in 0.1% PBS (wt/vol) casein were added and allowed to react for 30 minutes. Bound antibodies were detected by use of antihuman immunoglobulin G peroxidase conjugate (EUROIMMUN Medizinische Labordiagnostika AG) and stained with tetramethylbenzidine (EUROIMMUN Medizinische Labordiagnostika AG) for 15 minutes. All steps were performed at room temperature. The optical density was read at 450 nm using an automated spectrophotometer (Spectra Mini; Tecan, Crailsheim, Germany). A highly positive index patient serum was used to generate a standard curve consisting of three calibrators (10, 100 and 800 international units (IU)/ml). IU were calculated for all samples using this three-point standard curve. The cutoff was optimized either by receiver operating characteristic (ROC) curve analysis (maximal sum of sensitivity plus specificity) or by predefined specificities of 98% and 99%. Commercially available anti-dsDNA ELISA, antinucleosome ELISA, CLIF and Farr assays (all from EUROIMMUN Medizinische Labordiagnostika AG) were used as reference assays and were performed according to the manufacturer's instructions.

The global reactivity of Anti-dsDNA-NcX ELISA and the diagnostic significance of the tests were assessed by ROC curve analysis, and the areas under the curve (AUC) were calculated using GraphPad Prism 5 software (GraphPad, La Jolla, CA, USA). Statistical analysis regarding autoantibody test results and disease variables obtained from medical records were calculated using the Mann-Whitney U test in SPSS version 16 software (SPSS, Inc., Chicago, IL, USA). Correlation of global disease activity according to the modified Systemic Lupus Erythematosus Disease Activity Index (mSLEDAI 2000) 25 with antibody assay titers was calculated using Spearman's rank order correlation (r_s) test in GraphPad Prism 5 software. Linear regression analysis was used to assess the significance of correlations for changes in disease activity and biomarkers over time. P < 0.05 was considered statistically significant.

To assess the reactivity of the novel Anti-dsDNA-NcX ELISA, sera of 207 SLE patients, 400 healthy donors and 357 patients with different rheumatic diseases relevant in the differential diagnosis of SLE were measured (Figure 1A). At the manufacturer's threshold of 100 U/ml, a sensitivity of 60.4% and a specificity of 98.9% were calculated for the diagnosis of SLE. The results for eight disease controls were false-positives. Six of these eight non-SLE patients being positive in Anti-dsDNA-NcX had either positive anti-dsDNA or antinucleosome ELISA results at a specificity of 98.9% (Figure 1B). Notably, none of these eight SLE patients tested positive in the CLIF or Farr assay.

Out of interest, the medical records of all non-SLE patients who tested positive in the Anti-dsDNA-NcX ELISA were checked (Figure 1B). In five patients, causes other than SLE with the potential to induce anti-dsDNA antibodies were found, namely, drugs and fever indicating an unknown infection. Combining Anti-dsDNA-NcX test results and medical history with consequent exclusion of these five patients would lead to an increased specificity of 99.5% at the manufacturer's threshold.

To further assess the performance criteria of the Anti-dsDNA-NcX ELISA with those of other assays for measuring antibodies against dsDNA and/or nucleosomes, the new ELISA was compared with the anti-dsDNA ELISA, the antinucleosome ELISA (free of Scl-70 and histone H1) and the Farr assay in sera from 964 individuals by using ROC curve analysis (Table 1).

To check whether the performance of a single test system was significantly better than another one, we additionally tested the reported AUC values for significant differences. This formal statistical comparison revealed that the Anti-dsDNA-NcX ELISA and the antinucleosome ELISA were significantly better than the anti-dsDNA ELISA (p = 0.0024 and p = 0.0029, respectively). The Farr assay was not significantly better than any other ELISA, nor was the opposite the case.

The Anti-dsDNA-NcX ELISA revealed superior results in all performance criteria. The CLIF assay was not integrated into ROC curve analyses because it is a semiquantitative test. However, a sensitivity of 28.02% at a specificity of 98.15% was separately calculated for the CLIF assay. To allow direct comparison, the sensitivities of all other test systems are also shown at a specificity of 98.15% in Table 1.

As considerable differences were obtained in ROC curve analysis, two clinically relevant questions were addressed to further elucidate the similarities (question 1) and differences (question 2) of the test systems used. First, how often are sera positive in one test and positive in another test system? Second, how often are sera negative in one test but positive in another test system? Answers to the second question would give physicians clues to how often they might miss the correct diagnosis by determination of disease-specific autoantibodies using only one test system.

To integrate the CLIF assay with its specificity of 98.15%, individual cutoffs of the other test systems at the same specificity (see also Table 1) were used. Intersections of the three ELISAs (Anti-dsDNA-NcX, antinucleosome and anti-dsDNA) and separately of dsDNA-NcX with the Farr and CLIF assays are illustrated in a Venn diagram shown in Figure 2. Of 207 sera, 139 were positive in all of the three ELISAs, which were compared at cutoffs read out at a specificity of 98.15%. Of these 139 positive sera, 99.28% were positive with the Anti-dsDNA-NcX ELISA, while 9.28% were exclusively positive with the Anti-dsDNA-NcX ELISA. In a comparison of the different detection techniques (Figure 2B), 149 sera were positive in the Farr assay, Anti-dsDNA-NcX ELISA or the CLIF assay. Exclusively positive sera were found as follows: two (1.3%) in the CLIF assay, six (4.0%) in the Farr assay and 29 (19.5%) in the Anti-dsDNA-NcX ELISA. Among these three tests, 138 (92.6%) of all 149 positive sera were positive in the Anti-dsDNA-NcX ELISA.

To answer the second question and to reveal differences between the assays, we further determined how often serum is negative in one test but positive in another test (Table 2). Clinically relevant for the verification of diagnosis, these frequencies indicate how often a diagnosis of SLE may be missed by using only one test.

To reveal clinical associations of investigated test systems, assay titers of patients with a distinct present clinical finding were compared with those of patients without that finding using the Mann-Whitney U test (Table 3). Clinical findings were read out of medical records consisting of ACR criteria, mSLEDAI 2000 score, laboratory parameters and immunosuppressants and/or antimalarials.

Using the global mSLEDAI 2000 score (items used to score for anti-dsDNA and complements were excluded to avoid bias), only the Anti-dsDNA-NcX ELISA (r_s = 0.145, p = 0.034) and anti-Nuc ELISA (r_s = 0.143, p = 0.034) were significantly correlated on the basis of Spearman's rank order correlation Spearman's rank order correlation, but neither the anti-dsDNA ELISA (r_s = 0.113, p = 0.074) nor the Farr assay (r_s = 0.118, p = 0.065) showed a significant correlation.

All test systems were significantly associated with urinary casts and decreased complement component 3 (C3) at the time of blood sampling. Notably, some items were related to only one assay. So, the mSLEDAI 2000 items pericarditis and decreased complement component 4 (C4) were exclusively associated with elevated titers in the anti-dsDNA ELISA. The Farr assay was exclusively connected to the presence of pleuritis. The CLIF assay was associated with prior hematological manifestation according to an ACR criterion. It is noteworthy that the SLEDAI 2000 item proteinuria was solely associated with higher levels found by the Anti-dsDNA-NcX ELISA.

Of our 207 patients, 101 had histologically proven lupus nephritis. Of those 101 patients, 61 were not considered in further subgroup analysis because of a missing biopsy within 1 year before blood draw. Seven patients actually had lupus nephritis class II, 13 had class III, 12 had class IV and eight had class V according to the International Society of Nephrology Working Group on the Classification of Lupus Nephritis/Renal Pathology Society Working Group on the Classification of Lupus Nephritis guidelines 26. Prior or present renal involvement (n = 101) was significantly associated with elevated test results in the anti-NcX ELISA (P = 0.002), anti-Nuc ELISA (P = 0.004), anti-dsDNA ELISA (P = 0.008) and the Farr assay (P = 0.042), but not in the CLIF assay (P = 0.812) (all P values based on the Mann-Whitney U test). Further analysis of defined classes revealed an exclusive correlation of lupus nephritis class IV class (n = 12) with elevated antibody levels on the basis of Anti-Nuc, Anti-dsDNA and Anti-dsDNA-NcX ELISA results using Spearman's rank order correlation (r_s = 0.192, r_s = 0.189, r_s = 0.183, respectively; all p < 0.01).

Monitoring of disease activity is of prime importance in the clinical care of SLE patients. Regarding the correlation of anti-dsDNA antibodies with disease activity, contradicting reports have been published in the literature, ranging from a positive 8 over a missing 9 to a negative correlation 7. In the longitudinal prospective approach over 10 months used in our present study, 20 SLE patients were monitored with an overall total of 69 patient visits resulting in 49 different data points. In all visits, no change in therapy within the past 4 weeks had occurred before medical records and blood were taken. To test whether changes in distinct laboratory parameters can reflect changes in disease activity, the anti-dsDNA ELISA, the Farr assay, the antinucleosome ELISA, the Anti-dsDNA-NcX ELISA, C3 and C4 were compared with the changes in mSLEDAI 2000 score over time (Figure 3). The mSLEDAI 2000 score was defined with an exclusion of items for dsDNA and complement components to avoid bias. In the follow-up study, none of the traditional biomarkers (anti-dsDNA ELISA, Farr assay or C3 or C4) were correlated with disease activity over time. However, Anti-dsDNA-NcX ELISA correlated best (r = 0.4970, P = 0.0003), followed by the antinucleosome ELISA (r = 0.4605, P = 0.0009) using linear regression. A subgroup analysis was not performed because of the limited number of patients.