In this study, we collected non-duplicated and non-selected sera from 100 patients who underwent anti-ENA testing at one university hospital from December 2011 to January 2012. The appropriate sample size with which to evaluate the performance of the diagnostic tools was determined based on a previous report.14 Of the 100 patients, 60 patients were diagnosed with well-defined SARDs [25 SLE, 24 SjS, 5 rheumatoid arthritis (RA), 4 MCTD, and 2 SSc]; 17 were diagnosed with non-SARD autoimmune disease, such as Raynaud disease, autoimmune carpal tunnel syndrome, or autoimmune disease with no criteria met for an exact diagnosis; 22 suffered from non-autoimmune diseases, such as osteoarthritis, allergic rhinitis, and hypertensive heart disease; and 1 patient's diagnosis was not available. The diagnosis of each patient with autoimmune disease was determined by expert rheumatologists. We received ethical approval of this study from the Institutional Review Board of Severance Hospital (IRB No. 1-2019-0013).

We detected the most commonly measured ENA specificities, such as anti-SS-A/Ro, anti-SS-B/La, anti-RNP, anti-Sm, and anti-Scl-70. Table 1 lists the nature of the antigens in the two tests that were evaluated.

Microplate ELISA for detecting five different anti-ENAs were conducted using QUANTA Lite™ ELISA kits (INOVA Diagnostics, San Diego, CA, USA). Laboratory procedures were performed by expert technicians according to the manufacturer's instructions, and manual sample preparation and manual distribution on the microplate initiated the wet process. After the procedure, the optical density (OD) of each reaction was measured by an ELISA reader, and the levels of autoantibodies were calculated using the values of OD from the patients' sera and positive controls. According to the manufacturer's instructions, the results were reported in three categories: negative, weak positive, and positive. Semi-quantitative results of less than 20 units were classified as negative, 20–39 units as weak positive, and more than 40 units as positive. For statistical analyses, weak-positive results were considered positive results.

As previously described in the literature, the Phadia® 250 assay (Phadia, Uppsala, Sweden) utilizes polystyrene EliA™ wells precoated with targeted antigens (SS-A/Ro protein, SS-B/La protein, RNP protein, SmD3 peptide, and Scl-70 protein, respectively).15 The fluorescence emitted from the conjugate of the autoantibody was detected by a built-in fluorometer, and the estimated value was expressed in EliA U/mL. The results were classified in three categories: negative, equivocal, and positive. According to the cutoff values provided by the manufacturer, results less than 7 EliA U/mL were classified as negative, 7–10 EliA U/mL as equivocal, and more than 10 EliA U/mL as positive. For statistical analyses, equivocal results were considered positive results.

The statistical tests in this study were conducted using Analyse-it® 3.90.4 (Analyse-it Software, Ltd., Leeds, UK) and R 3.5.3 (R Foundation for Statistical Computing, Vienna, Austria). Agreement between qualitative results from the two assays was assessed using kappa coefficients. Recommended interpretations of kappa values were as follows: poor agreement, 0.01–0.20; fair agreement, 0.21–0.40; moderate agreement, 0.41–0.60; substantial agreement, 0.61–0.80; and almost perfect agreement, 0.81–0.99.16 Degrees of association between quantitative values from the two tests were assessed using Spearman's correlation coefficients. A p value less than 0.05 was considered statistically significant. Sensitivity and specificity for each test were determined with two-by-two contingency tables. To define true positive, true negative, false positive, and false negative, a true event was defined as diagnosis of the patient with correlated disease. The sensitivities and specificities for the two methods were compared by estimation of confidence intervals for differences in paired sensitivities and paired specificities.1718 If the confidence limits for the differences in sensitivities or specificities did not include zero, there was evidence that the sensitivities or specificities were statistically different.18 As the clinical data of one patient was not available, analyses of diagnostic sensitivities and specificities were conducted with results from 99 patients.

The demographic data for the 99 patients are detailed in Table 2. Fifty-eight SARD patients were female (58/60, 96.7%), and 29 non-SARD patients were female (29/39, 74.4%). The average age of the SARD patients was 37.0 years, with a standard deviation (SD) of 17.3, and in non-SARD patients, the average age was 47.5 years, with a SD of 17.8. Sixty SARD patients comprised five different clinical diagnoses: SLE (n=25), SjS (n=24), RA (n=5), MCTD (n=4), and SSc (n=2). Non-SARD patients were subdivided into a non-SARD autoimmune disease group (n=17) or non-autoimmune disease group (n=22).

Overall agreement values between Phadia® 250 and ELISA assay are listed in Table 3. Based on values of concordant and discrepant results, the agreement rates between ELISA and Phadia® 250 ranged from 89% for anti-RNP to 97% for anti-Scl-70. The estimated kappa coefficients for agreement between the results by the two assays had a minimum value of 0.44 for anti-Sm and a maximum value of 0.82 for anti-SS-B/La. In detection of anti-Scl-70 and anti-Sm, the two methods showed moderate agreement with kappa coefficients of 0.56 and 0.44, respectively. For anti-SS-A/Ro and anti-RNP, the two methods demonstrated substantial agreement. Correlation of signal to cut-off ratios was analyzed with Spearman's rank correlation coefficients to document the degree of association between the two tests (Fig. 1). Spearman's coefficients between the results by the two methods were 0.93 for anti-SS-A/Ro, 0.72 for anti-SS-B/La, 0.43 for anti-RNP, 0.33 for anti-Sm, and 0.33 for anti-Scl-70.

The sensitivities and specificities of ELISA and Phadia® 250 in the detection of each anti-ENA antibody are shown in Table 4. As stated in the Materials and Methods section and Table 4, a true event in the evaluation of the diagnostic sensitivity and specificity of anti-SS-A/Ro was a diagnosis with SLE or SjS (total n=49). The rest of the diagnostic accuracy criteria for each autoantibody are detailed in the Table 4. According to the Clinical and Laboratory Standards Institute guideline EP12-A, we used 95% confidence intervals to determine statistically different differences.18 In general, in the provided clinical context in Table 4, more false positive cases were observed with the ELISA assay. Phadia® 250 showed higher sensitivity and specificity for the detection of anti-SS-A/Ro and anti Scl-70, although the difference was not statistically significant. Overall, there were no significant differences between the two assays, except for the specificity of anti-RNP and sensitivity of anti-Sm.