Serum complement 3 is a potential biomarker for assessing disease activity in Takayasu arteritis

A prospectively ongoing observational cohort—the East China Takayasu arteritis (ECTA) cohort—was established since 2010 centred at Zhongshan Hospital, Fudan University, Shanghai, China. All the registered patients in ECTA cohort were diagnosed as TA by experts according to the American College of Rheumatology (ACR) 1990 classification criteria [13]. Patients’ information was collected with a standardized form and stored at the Redcap database once diagnosed.

In the current study, baseline information including clinical characteristics, biomarkers, and medications were used for analysis. Patients complicated with the following diseases were excluded: (i) malignant tumour; (ii) acute infections (e.g. tuberculosis, hepatitis); and (iii) other autoimmune diseases (e.g. SLE, AAV). The baseline data of patients enrolled from April 2008 to June 2019 was extracted as the original dataset, while an independent group of patients from July 2019 to Dec 2019 were obtained as the validation dataset. The flowchart of study is listed in the Supplementary Fig. S1.

The investigation protocol was conformed with the Helsinki Declaration and approved by the Ethics Committee of Zhongshan Hospital, Fudan University (Approval No.: B2016-168). All the patients signed the informed consents prior to the enrolment.

Blood samples at baseline of each patient were obtained and examined in the clinical laboratory of the hospital with standard operating procedure. In detail, ESR was detected by Westergren method; CRP as well as globulin and immunoglobulin was detected by automatic biochemical analyser; and cytokines were detected by chemiluminescent immunoassay (CLIA) using Semen platform. Serum complements were detected by immunity transmission turbidity (ITA) in routine procedures according to the instructions of the manufactures with automatic biochemical analyser.

Kerr criteria were used as the gold standard for disease activity evaluation: (i) systemic symptoms (infection, tumour, etc., were excluded); (ii) elevated ESR levels; (iii) vascular ischemic symptoms or signs (weakened pulse or pulselessness, vascular bruits, or asymmetric blood pressure); and (iv) positive imaging results. New onset or worsening of two or more criteria indicated “active disease” [14]. Whole body enhanced magnetic resonance angiography (MRA) was performed instead of the traditionally used angiography specified in the Kerr criteria in each patient. Imaging types were identified according to the angiographic classification of the international TA conference in Tokyo (1996) based on lesion distribution [15].

Continuous variables are expressed as the mean ± standard deviation or median (interquartile range, IQR) and compared using the t test or Wilcoxon rank-sum test where appropriate. Categorical variables are presented as number (frequency) and compared with the chi-square test. The Spearman correlation analysis was employed to validate the relationship between C3 and other biomarkers. The univariate logistic regression analysis was executed to unearth the factors associated with disease activity, which was further observed reversely by cluster analysis with K-means method and principal component analysis based on the optimal number of groups determined by scree plots. Items with P < 0.05 in univariate logistic regression analysis were enrolled in the multivariate logistic regression analysis and performed the receiver operating characteristic (ROC) curves. Youden Index was employed to determine the optimal cut-off value and the corresponding diagnosis ability. The parallel test and serial test were further performed to evaluate the value of C3 in combined tests. Next, net reclassification index (NRI) and integrated discrimination index (IDI) were calculated to confirm the value of C3 in improving the ability of CRP or ESR in distinguishing active TA. In addition, the identification effect of the parallel test (CRP and C3) was validated by 10-fold cross-validation and external validation with the original dataset and the independent dataset respectively. A P < 0.05 was considered to be significant with a two-side test. The data were analysed using SPSS 22.0 (Chicago, IL, USA) and R software (Murray Hill, NJ, USA). The graphs were generated with Prism GraphPad 8.0 (San Diego, CA, USA) and R software.

In total, 519 patients were enrolled. Of these, 427 cases (82.3%) were female. The mean age was 32 (24–45) years, and the median disease course was 21 (4.0–70.5) months. The most common manifestation was dizziness (41.4%), followed by fatigue (33.3%) and chest pain (22.2%). Pulselessness was observed in 41.8% of patients, vascular bruit in 44.5% cases, and both were more commonly observed in the active group. The most common imaging type was type V (35.1%), followed by type I (28.5%), type IV (10.2%), type IIb (7.7%), type IIa (7.1%), and type III (3.9%) (Table 1).

According to Kerr criteria, 406 cases (72.2%) were defined as active disease. There was no significant difference between active and inactive groups in age, sex, disease course, and imaging types. Patients with active disease had higher frequency of fatigue, night sweat, chest pain, abdominal pain, pulselessness, and vascular bruit than patients with inactive disease (P < 0.05). The prednisone, cyclophosphamide, and methotrexate were the most commonly used drugs in TA patients, but there was no significant difference between two groups (Table 1).

Compared with those in inactive group, complements including C3 [1.00 (0.91–1.15) vs 1.19 (1.03–1.56), P < 0.001], C4 [0.22 (0.19–0.27) vs 0.25 (0.20–0.30), P = 0.040], and hemolytic complement (CH50, 59.3 ± 17.9 vs 66.4 ± 18.1, P = 0.015) were significantly higher in the active group. Moreover, serum levels of ESR, CRP, interleukin-6 (IL-6), platelets (PLT), globulin, immunoglobin A (IgA), and immunoglobin G (IgG) were also significantly higher in the active group compared with that in the inactive group (P < 0.05) (Table 1).

Then, logistic analysis was performed to clarify the relationships between different biomarkers with disease activity. In univariate regression analysis, it revealed that C3 and CH50, together with PLT, globulin, CRP, IgA, IgG, and IL-6, were positively correlated with disease activity (OR > 1, P < 0.05). However, in multivariate regression analysis, C3 levels [odds ratio [OR] (95%CI) 10.710 (1.825–62.835), P = 0.009] and CRP [OR (95%CI) 1.041 (1.009–1.073), P = 0.011] were independently associated with active disease (Supplementary Table S1).

In the scatter plot, C3, C4, and CH50 increased with Kerr score significantly and C3 was significantly positively associated with ESR, CRP, and IL-6 (r > 0.4, P < 0.001) (Fig. 1, Supplementary Fig. S2–3). Furthermore, cluster analysis showed that patients could be clustered into two groups based on C3, CH50, PLT, globulin, CRP, IgA, IgG, IL-6, ESR, and Hb levels. The principal component analysis revealed that C3, CH50, CRP, IL-6, PLT, and ESR were assigned to the same major component consisting of inflammatory biomarkers (Fig. 2). These data indicated that C3 levels were associated with the disease activity.

In the diagnosis test, the C3 cut-off value for identifying active disease was 1.085 g/L, with the sensitivity of 69.9%, the specificity of 66.7%, and the AUC of 0.715 (0.650–0.781). In comparison, the cut-off of CRP was 10.65 mg/L, with sensitivity of 50.7%, specificity of 82.4%, and the AUC of 0.703 (0.647–0.760). Moreover, the cut-off of ESR was 26.5 mm/H, with 63.8% sensitivity, 73.1% specificity, and the AUC of 0.766 (0.698–0.800). Further combined tests showed that combining CRP and C3 in parallel test, the sensitivity could be improved to 85.1%and the specificity was 55.0%. In serial test, the specificity was improved to 94.1%, with a sensitivity of 35.4% (Table 2, Fig. 1).

In addition, results of the NRI [OR (95%CI) 0.328 (0.224–0.431), P < 0.001] and IDI [OR (95%CI) 0.389 (0.312–0.466), P < 0.001] revealed that the introduction of C3 improved the ability of CRP to distinguish disease activity significantly.

In the internal validation, the 10-fold cross-validation revealed that the accuracy of the training group was more than 75% in the parallel test of CRP and C3, among which the maximal accuracy was 90.9% in the training group with the corresponding accuracy of 83.7% in the test group (Supplementary Table S2).

In the external validation, 53 cases of independent TA patients were employed to further validate the universality of the results, among whom, 24 (55.8%) cases were in active status according to the Kerr criteria. There was no significant difference between patients from the original dataset and validation dataset in age, sex, and levels of ESR, CRP, and C3. The AUC of CRP and C3 was 0.721 and 0.692, with the accuracy of 72.0% and 67.3%, respectively in validation dataset. The AUC was 0.721 with the accuracy of 72.7% in the parallel test of CRP and C3, while the AUC of serial test of CRP and C3 was 0.721 with the accuracy of 70.4%. The scatter plot also showed that the introduction of C3 could reclassify some patients with CRP below the cut-off value as active disease (Fig. 3, Supplementary Table S3–4).