Sample demographics
Demographic characteristics of the cohort are presented in Table 1. For juvenile patients, the study population included 221 continuous cases with a well-established clinical diagnosis of jSLE (83.7% female, a mean age at diagnosis of 10.9 ±2.7 years) and 214 controls. The jSLE group had a higher female ratio (P < 0.001) and older diagnosis age (P < 0.001) than the controls. In adult population, 221 cases with aSLE (90.5% female, a mean age at diagnosis of 36.8 ±12.8 years) and 214 controls were enrolled in this study. The aSLE had a higher female ratio (P < 0.001) and younger diagnosis age (P < 0.001) than the controls. Both jSLE and aSLE group had a higher proportion of patients with positive ANA than the control group. The control groups included a variety of diseases detailed in Table 1. In juvenile patients, juvenile idiopathic arthritis (n=90) and vasculitis (n=44) were the most common. In adult patients, the most prevalent diagnoses were dermatomyositis (n=30) and vasculitis (n=30).
Performance of the criteria
Performance of EULAR/ACR-2019 criteria, SLICC-2012, and ACR-1997 are presented in Table 2. In jSLE patients, the sensitivity of the EULAR/ACR-2019 criteria was the highest (98.2%), while the highest specificity was that of the ACR-1997 criteria (99.5%). The SLICC-2012 criteria performed well with high sensitivity and specificity (94.6% and 98.6%). For aSLE, both the EULAR/ACR-2019 and SLICC-2012 criteria had higher sensitivity (99.1% and 96.8%, respectively) than the ACR-1997 (72.9%), while the ACR-1997 criteria had higher specificity than the other two sets (97.2% vs. 90.2%–92.5%).
Performance of the criteria in ANA positive patients
Since a positive ANA is the entry criteria in EULAR/ACR-2019 criteria, we also evaluated the performances of the three criteria sets by including only ANA positive SLE patients and controls (Table 2). There were three ANA negative patients in our jSLE cohort, the sensitivities of the ACR-1997, SLICC-2012 and EULAR/ACR-2019 citeria were 64.2%, 94.5%, and 99.5%, respectively; while the specificities were 98.3%, 94.8%, and 75.9%, respectively. For adult patients, the sensitivities of the ACR-1997, SLICC-2012 and EULAR/ACR-2019 criteria were 72.9%, 96.8%, and 99.1%, respectively. The specificities in aSLE patients were 95.7%, 88.7%, and 85.1%, respectively. When including only ANA positive SLE patients and controls, we noticed a difference in specificities by the EULAR/ACR-2019 criteria. Especially in jSLE patients, the specificity of EULAR/ACR-2019 criteria decreased from 93.5% (all patients) to 75.9% (including only ANA positive patients).
Receiver operating characteristic (ROC) analysis showed a good performance of EULAR/ACR-2019 criteria in discriminating SLE from other rheumatic disease patients (Figure 1). Among ANA positive juvenile patients, the area under the ROC curve for EULAR/ACR-2019 criteria was 0.976 (95%CI, 0.959-0.994), and the cut-off value was 13, resulting in a sensitivity and a specificity of 92.2% and 93.1%, respectively. When we picked ≥13 as the threshold for the EULAR/ACR-2019 criteria, the sensitivity slightly decreased (from 99.5% to 92.2%), but the specificity significantly increased (from 75.9% to 93.1%). For adult patients, the area under the ROC curve for EULAR/ACR-2019 criteria was 0.965 (95% CI, 0.946-0.985), and the cut-off value was 10 (sensitivity, 99.1%; specificity, 85.1%).
Proportion of patients meeting overlapping or individual criteria
Among the clinically diagnosed SLE patients, 140 (63.3%) jSLE and 161 (72.9%) aSLE patients met all three classification criteria. There were twelve jSLE patients and seven aSLE patients who only met the EULAR/ACR-2019 criteria, of whom eleven and four cases had single organ involvement, respectively. Four cases of jSLE patients were missed by the EULAR/ACR-2019 criteria, three of whom were ANA negative (one turned ANA positive three months after diagnosis). Two aSLE patients were missed by the EULAR/ACR-2019 criteria, both were ANA positive, and one patient got a EULAR/ACR total score of eight (kidney biopsy showed V-type lupus nephritis) (Figure 2). Fourteen children and twenty-one adult patients in controls were misclassified as SLE by the EULAR/ACR-2019 criteria. Patients with undifferentiated connective tissue disease (UCTD) and mixed connective tissue disease (MCTD) were most likely to be classified as SLE by the EULAR/ACR-2019 criteria in our cohort.
Sensitivity of the criteria across subgroups of SLE patients with major organ involvement
SLE patients with major organ involvement often accrue more damage, which has negative impact on long term survival. The prompt diagnosis and management is essential to improve patients’ prognosis. Thus, we specifically analyzed the diagnostic performance of the three criteria in patients with major organ involvement. Compared with the other two classification criteria, ACR-1997 had lower sensitivity for classifying neurological, renal, and hematological SLE in both children and adults. The SLICC-2012 and EULAR/ACR-2019 criteria had increased sensitivity for hematological manifestations. In jSLE patients, the sensitivity of the SLICC-2012 and EULAR/ACR-2019 criteria were the same (97.6%), while in aSLE, the sensitivity of EULAR/ACR-2019 was slightly higher (99.3% vs. 97.9%). The sensitivity of the SLICC-2012 for classifying renal jSLE was slightly higher than EULAR/ACR-2019 criteria (100% vs. 96.9%), whereas in aSLE, the sensitivity of the two criteria sets was the same (99.2%). The three criteria set had high sensitivity for classifying SLE patients with moderate and severe activity (SLEDAI-2K ≥10) (Figure 3).
jSLE patients fulfilling the EULAR/ACR-2019 criteria had a mean score of 19.3 (range:10-33). SLEDAI-2K ranged from 3 to 26 (mean, 10.2). aSLE patients who met the EULAR/ACR-2019 criteria scored 10-41 (mean, 20.8). SLEDAI-2K ranged from 2 to 27 (mean, 12.3). There was a positive correlation between EULAR/ACR total score and SLEDAI in jSLE and aSLE groups (r=0.700, P < 0.001; r=0.748, P < 0.001) (Figure 4).