This single-center, retrospective study screened serum samples from consecutively submitted laboratory samples at Peking Union Medical College Hospital between July 2022 and July 2023. Patients were eligible if they fulfilled the 1997 ACR criteria (6), the 2012 Systemic Lupus International Collaborating Clinics (SLICC) classification criteria (23) or 2019 EULAR/ACR criteria (24). Additionally, patients diagnosed with Sjögren’s syndrome (SS) according to the 2016 ACR/EULAR classification criteria (25), or rheumatoid arthritis (RA) according to the 2010 ACR/EULAR classification criteria (26) were included as disease control groups. In brief, 3081 patients were enrolled, comprising 1509 SLE patients, 662 SS patients, and 910 RA patients. A total of 3429 serum samples were collected, including 1773 from SLE patients, 937 from RA patients, and 719 from SS patients.

Clinical information at baseline was extracted from patient medical records, including age, sex, antinuclear antibodies (ANA) titer and fluorescence pattern, and other laboratory abnormalities. Moreover, 351 SLE patients had detailed clinical assessments which had been reported previously (27), including organ manifestations and SLEDAI-2000 score. This study was approved by the Medical Ethics Committee of PUMCH (Beijing, China) and written informed consent was obtained from all participants before inclusion in the study.

Serum samples were collected from all patients and stored at –80°C and were thawed only once before testing to preserve antibody integrity. The anti-dsDNA IgG antibodies were measured using three different immunoassays performed according to the manufacturer’s instructions.

Based on the manufacturer’s recommended cut-off values, sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV) were calculated for each assay and their respective combinations. McNemar’s test was utilized to compare the paired proportions between the diagnostic methods. Receiver operating characteristic (ROC) curves were generated to evaluate the diagnostic accuracy of each assay, and the area under the curve (AUC) was calculated. The inter-method agreement was assessed using Cohen’s kappa coefficient (κ), which accounts for agreement occurring by chance. Pearson correlation analysis was used to explore the correlation between anti-dsDNA concentration and SLEDAI-2k scores. Comparisons between groups were made using the Mann-Whitney U test or the Wilcoxon signed-rank test, as appropriate. A two-tailed p-value of <0.05 was considered statistically significant. Statistical significance was defined as a two-tailed p-value of <0.05. All statistical analyses were performed using SPSS software (version 26.0.0.0; IBM Corp., Armonk, NY, USA) and Python (version 3.11.8) with the sklearn.metrics module.

A total of 3081 patients were included in the analysis, comprising 1509 patients with SLE and 1572 other autoimmune disease controls (662 with SS and 910 with RA). The baseline demographic and clinical characteristics of the study population are summarized in Table 1 . The mean age of patients with SLE was 33.98 ± 14.44 years, which was significantly younger than that of the control group (48.19 ± 14.28 years; p < 0.05). A female predominance was observed across all groups, consistent with the epidemiology of autoimmune diseases. Among SLE patients, 96.62% tested positive for ANA, predominantly exhibiting speckled (70.6%) or homogeneous (64.6%) patterns. The ANA titers across SLE patients are presented in Supplementary Table 1 . Furthermore, the prevalence of anti-RNP, anti-Sm and anti-rRNP antibodies was significantly elevated in SLE patients (p<0.001).

Among the 351 SLE patients with detailed clinical assessments, 398 samples were analyzed. Organ involvement and disease activity were presented in Table 2 . The mean age of the 351 patients was 34.46 ± 11.86 years, which was comparable to the overall SLE cohort. Cutaneous (281, 70.60%) and hematologic (257, 64.57%) manifestations were most frequent. Renal involvement was observed in 143 samples (35.93%). Disease activity, as assessed by SLEDAI-2k scores, was predominantly mild (score ≤6) in 302 samples (75.88%). Moderate activity (score 7–12) and severe activity (score >12) were observed in 65 (16.33%) and 31 samples (7.79%), respectively.

The diagnostic performance of individual and combined anti-dsDNA antibody detection methods is presented in Table 3 and Supplementary Table 2 . All three methods performed high specificity, as depicted in Figure 1 . The IIF assay demonstrated the highest specificity (98.31%) and PPV (96.10%); however, its sensitivity (38.92%) was significantly lower compared to DLCM (43.65%) and CLIA (41.57%) (p < 0.05). DLCM exhibited the highest sensitivity and accuracy (69.29%) with a specificity of 96.7%. NPVs for all three methods were relatively similar, ranging from 60.05% to 61.59%.

In contrast, combining two assays led to a statistically significant improvement in sensitivity compared to any single assay alone. As shown in Table 3 , the combination of IIF and DLCM resulted in a sensitivity of 52.23% (p < 0.001 vs. IIF) and a specificity of 95.89%. The IIF/CLIA combination followed closely with a sensitivity of 50.99% and an accuracy of 71.95%. Notably, combining all three assays increased sensitivity to 56.40%, while specificity slightly decreased to 93.72%.

ROC curve analysis confirmed these findings. The AUC for DLCM was 0.71 (95% CI [0.68, 0.73]), indicating moderate diagnostic accuracy, while AUCs for IIF and CLIA were both 0.69. Combining IIF and DLCM increased the AUC to 0.76, and the combination of all three assays yielded the highest AUC of 0.77 (95% CI [0.75, 0.79]), underscoring the diagnostic advantage of assay combinations.

Various patterns of assay positivity were observed and the distribution of positivity across the three detection methods is detailed in Table 4 and illustrated in Figure 2 . Among the 1773 SLE samples, 467 (26.3%) tested positive for anti-dsDNA antibodies by all three methods, whereas only 10 control subjects (0.6%) exhibited triple positivity. Additionally, 96 patients (5.4%) were positive exclusively by IIF, another 96 (5.4%) solely by DLCM, and 74 (4.2%) only by CLIA. These patterns imply that each assay would capture distinct subsets of anti-dsDNA antibodies. Combining multiple detection methods enhances the overall sensitivity and diagnostic yield for identifying patients with systemic lupus erythematosus. The Venn diagrams in Figure 2 depict the overlap of positive results among the three assays for both SLE patients and control subjects.

The concordance between the detection assays was assessed using Cohen’s kappa coefficients. There was substantial agreement between DLCM and CLIA (κ = 0.78), indicating a high level of consistency in their detection of anti-dsDNA antibodies. In contrast, the agreement between IIF and DLCM was moderate (κ = 0.65), as was the agreement between IIF and CLIA (κ = 0.66). These results suggest that while DLCM and CLIA have similar detection capabilities, IIF may identify more different antibody subsets.

Figure 3 illustrates the hierarchical distribution of anti-dsDNA antibody levels detected by the three assays in relation to the clinical diagnoses. The assay results were stratified into grades from 0 to 4 based on specific concentration ranges to facilitate comparison across methods. Analysis revealed that higher-grade antibody levels (grades 3 and 4) were more frequently observed in CLIA and DLCM compared to IIF (p < 0.001). Specifically, the proportion of patients with high antibody titers was significantly greater in CLIA and DLCM assays than in the IIF assay, indicating that CLIA and DLCM are more sensitive in detecting elevated levels of anti-dsDNA antibodies.

To evaluate the consistency of the detection methods in reflecting changes in anti-dsDNA antibody levels over time, we analyzed data from 88 SLE patients who underwent two serial anti-dsDNA antibody tests using all three assays during the study period. For each patient, the change in antibody titer grade between the two tests was calculated for each method by subtracting the initial grade from the subsequent grade, based on predefined grading criteria specific to each assay (as previously described).

Figure 4 illustrates the distribution of titer grade changes for each assay. The majority of patients exhibiting concordant grade increases or decreases between consecutive tests. However, CLIA and DLCM more consistently detected changes in antibody levels. In contrast, the IIF method displayed greater variability. CLIA and DLCM might track antibody changes more reliably in SLE monitoring

Among 398 SLE samples with detailed clinical information, the association of anti-dsDNA antibody levels detected by each assay with SLE disease activity was evaluated ( Figure 5 ). Disease activity was quantified using the SLEDAI-2k score. The median anti-dsDNA titers assessed by IIF was negative (IQR [-,1:10]) in mild patients (SLEDAI-2k ≤ 6), and 1:10 (IQR [-,1:20]) in medium and severe patients. Pearson’s correlation analysis revealed a moderate positive correlation between anti-dsDNA concentrations measured by DLCM and SLEDAI scores (R = 0.4212, p < 0.001). Similarly, CLIA-derived anti-dsDNA levels showed a positive correlation with SLEDAI scores, though the association was weaker than that observed with DLCM (R = 0.2852, p = 0.015).

Previous studies have reported an association between anti-dsDNA levels and renal involvement. Among the 398 SLE samples with detailed clinical information, 143 were from patients with renal involvement and 255 from those without ( Table 5 ). Anti-dsDNA levels were compared between the two subgroups ( Figure 6 ). Both IIF and CLIA methods showed significant differences between the subgroups (p < 0.05 and p < 0.001, respectively). However, quantitative DLCM measurements revealed no statistically significant difference in anti-dsDNA concentrations between the renal and non-renal groups (p = 0.132).

The sensitivity of the three methods in the renal involvement group did not show significant differences, with IIF demonstrating the highest specificity (69.02%) ( Table 4 , Supplementary Table 3 ). Compared to single methods, dual combinations significantly improved sensitivity (e.g., IIF/DLCM 56.64% vs IIF 48.95%, p = 0.001). However, the specificity of the combinations decreased (e.g., IIF/DLCM 53.73% vs IIF 69.02%, p < 0.001), highlighting a trade-off between early detection and false positives.