Introduction
Autoimmune diseases (AIDs) are associated with specific mechanisms that may predispose to arterial and venous thromboembolism [1], [2]. These disorders are characterized by immune dysregulation, in which the immune system mistakenly attacks self-tissues, resulting in chronic inflammation. This inflammation promotes a hypercoagulable state through endothelial dysfunction, platelet activation, release of procoagulant microparticles from platelets, leukocytes, and endothelial cells, tissue factor expression, inhibition of the protein C system, and impaired fibrinolysis [3], [4], [5].
In a comprehensive study involving 535,538 patients hospitalized for autoimmune diseases, Bengt Zöller and colleagues demonstrated that 33 specific autoimmune diseases are notably linked to an increased risk of pulmonary embolism within the first year following hospitalization [6]. Systemic lupus erythematosus (SLE) and primary Sjögren's syndrome (pSS) are significant autoimmune disorders [7]. SLE is a condition capable of impacting various organ systems, such as the skin, kidneys, blood vessels, and the nervous system [8].
Thrombosis is commonly seen in SLE and plays a crucial role in increasing morbidity and mortality rates [9]. In individuals diagnosed with SLE, arterial thromboembolism occurs in an estimated 5.1% to 8.5% of cases within the first five years after diagnosis. Venous thrombosis rates have also been rising, now affecting somewhere between 3.7% and 10.3% of patients [10]. pSS is a type of inflammatory condition marked by the accumulation of mononuclear cells within exocrine glands, leading to dryness in the mouth and eyes [11].
A previous study highlighted that through the first year of diagnosis of SS, venous thromboembolism risk is seven times more than matched healthy controls [12]. In addition, a meta-analysis indicated that the combined risk ratio for venous thromboembolism linked to pSS was 2.05 (95% CI, 1.86-2.27) [13]. Figueroa et al. [14] have directly compared the comorbidity burden between pSS and SLE in a Spanish cohort, demonstrating that pSS patients have a lower prevalence of classic cardiovascular risk factors (smoking, dyslipidaemia, hypertension) and of life-threatening cardiovascular events (stroke or myocardial infarction) than SLE patients.
However, whether the risk-factor profiles for arterial and venous thrombosis differ directly between SLE and pSS, particularly in Asian populations, remains largely unexplored.
This study utilized a large single-center Chinese cohort with extended follow-up to systematically compare the incidence and risk-factor profiles of overall, arterial, and venous thrombosis in patients with SLE and pSS. Specifically, we aimed to estimate and directly compare the cumulative incidence of overall, arterial, and venous thrombosis between the two diseases, to identify and compare the independent risk factors associated with each thrombotic subtype using multivariable Cox proportional hazards regression, and to develop and internally validate disease-specific risk prediction models for thrombotic events. The present study represents the first comprehensive, long-term characterization of the thrombotic burden and its determinants in a large Asian cohort, and is expected to provide a basis for risk stratification and targeted thromboprophylaxis in Chinese patients with SLE and pSS.
Materials and methods
Study participants
A retrospective analysis was conducted for patients (aged ≥18 years) who were diagnosed with autoimmune disorders and admitted to our hospital between 2010 and 2024. The diagnosis of SLE in patients was established by rheumatologists using the SLE classification criteria set forth by the European League Against Rheumatism and the American College of Rheumatology in 2019 [8]. For patients with pSS, the diagnosis was made according to the 2016 classification criteria from the American College of Rheumatology/European League Against Rheumatism [15].
This study included a total of 955 patients with SLE and 558 patients with pSS. The date and type of every thrombotic event were recorded. Patients with incomplete data on follow-up dates, thrombotic events, demographics, treatment history, disease characteristics, or conventional thrombotic risk factors were excluded.
Patients who experienced a thrombotic event prior to the initiation of follow-up were also excluded.
Demographic and disease features
Demographic and clinical characteristics: sex, age at disease onset, time to thrombosis, and history of lupus nephritis.
Treatment history: prior use of hydroxychloroquine at any point.
Traditional risk factors: elevated body mass index (>25 kg/m2), diabetes, hypertension, hyperlipidaemia, smoking history.
Disease-related laboratory markers: positive anti-double-stranded DNA antibodies (dsDNA), reduced complement levels (C3 and C4), elevated platelet count (>350 × 109/L), positive D-dimer (>0.5 mg/L fibrinogen equivalent units [FEU]), and positive antiphospholipid antibodies (aPLs). The aPLs panel comprised lupus anticoagulant (LA), anti-β2-glycoprotein I (aβ2-GP1), and anti-cardiolipin antibodies (aCLs).
APLs were strictly defined according to the 2023 ACR/EULAR classification criteria for antiphospholipid syndrome [16]. LA was deemed positive when the normalized dRVVT ratio (screen ratio divided by the confirmation ratio) exceeded 1.2 in accordance with ISTH recommendations [17]. LA testing was conducted twice, with a 12-week interval between them.
Only those cases with positive results from both tests were included in the study. Mixed experimental findings indicate that false positives have been effectively eliminated. The potential causes of false positives include the use of anticoagulants, the absence of common pathway coagulation factors, and the presence of inhibitors to common pathway coagulation factors. aCLs and aβ2-GP1 were considered positive (≥40 U).
All positive aPL results were confirmed on a second determination ≥12 weeks later.
Definition of thrombosis
Venous thromboembolism (VTE) and arterial thromboembolism (AT) refer to any thrombotic incidents arising following the diagnosis of SLE or pSS. Confirmation of these incidents is achieved through imaging techniques, including Doppler ultrasound, computed tomography angiography, or magnetic resonance imaging. VTE includes deep vein thrombosis (DVT), pulmonary embolism (PE), and thrombosis affecting various venous systems, such as the portal and mesenteric.
AT encompasses events such as cerebral infarction, myocardial infarction, and peripheral arterial thrombosis, which includes renal artery and mesenteric artery thrombosis.
Statistical analysis
The statistical analyses were conducted using GraphPad Prism 9 (GraphPad Software Inc., USA) along with SPSS 26.0 software (SPSS Inc., USA). For continuous variables, medians and interquartile ranges (IQRs) were utilized, while categorical variables were presented in terms of percentages.
The follow-up period began with the diagnosis of the autoimmune condition and concluded upon the occurrence of the first thrombotic event, patient dropout, or death, depending on which came first. Patients without a thrombotic event were censored at the last follow-up visit. To explore the relationship between each variable and thrombotic events, a univariate Cox proportional hazards regression analysis was initially conducted.
Variables that demonstrated a P-value of less than 0.15 in the univariate analysis were later included as covariates in the multivariable Cox proportional hazards models. Variable selection was informed by expert consensus and published evidence and included: age at disease onset, sex, smoking history, history of lupus nephritis, hydroxychloroquine treatment, diabetes, hypertension, hyperlipidaemia, and the laboratory markers: reduced complement levels, positivity for anti-dsDNA antibodies, elevated platelet count, positive D-dimer, and positivity for aPLs.
The prevalence of each risk factor and the incidence of thrombosis were compared between the SLE and pSS cohorts using the chi-square test or Fisher's exact test for categorical variables and the Mann-Whitney U test for continuous variables. A P value < 0.05 was considered statistically significant.
To further identify variables specifically associated with different types of thrombosis, supplementary analyses were conducted using the same univariate-multivariable Cox regression framework, with venous and arterial thrombosis as separate outcome events. A P < 0.05 threshold indicated statistical significance. The Kaplan-Meier method was utilized to estimate survival curves.
A prediction model was built using the selected independent predictors. To further quantify the incremental predictive value of individual risk factors, we calculated the net reclassification improvement (NRI) for each independent predictor, comparing models with and without the factor of interest. The individual risk score was derived from the linear predictor of the multivariable Cox model.
For each landmark time (3, 5, and 10 years), the optimal risk-score cutoff was selected by maximizing the Youden index on the time-dependent receiver operating characteristic (ROC) curves.
To comprehensively evaluate the clinical practicability and diagnostic efficacy of the prediction models beyond conventional discriminative metrics, positive likelihood ratio (LR+) and negative likelihood ratio (LR−) were calculated at the Youden-derived optimal cutoff values of the risk score. The 95% confidence intervals of all LR values were rigorously estimated via 1000 bootstrap resampling iterations with replacement to avoid sampling bias and ensure statistical robustness. The evaluation of discrimination was conducted using the time-dependent area under the curve (AUC) with bootstrap bias correction and Harrell's C-index.
Results
Participant characteristics
Initially, 1049 patients diagnosed with SLE and 636 patients diagnosed with pSS were screened. Reasons for exclusion included imprecise follow-up start or end dates, incomplete data, and thrombotic events occurring before the diagnosis of the autoimmune disease. After excluding 94 SLE patients and 78 pSS patients, 955 patients with SLE and 558 with pSS remained for the final analysis.
Median follow-up was 10 years (IQR 8-12) in SLE and 9 years (IQR 7-12) in pSS, with the longest follow-up period extending to 15 years.
According to Table 1, patients with pSS were older than those with SLE at both disease onset and the occurrence of thrombotic events (P < 0.05). The two groups also exhibited significant differences in their profiles of traditional cardiovascular risk factors: SLE patients had a markedly higher prevalence of smoking (10.7% vs. 6.6%, P = 0.01), hypertension (25.9% vs.
21.0%, P = 0.032), and hyperlipidaemia (13.6% vs. 6.8%, P = 0.001), while diabetes was more prevalent among pSS patients (4.1% vs. 7.5%, P = 0.006).
In terms of laboratory markers, the prevalence of low complement levels significantly differed between the two groups (85.9% vs. 58.6%, P = 0.002). Concerning disease-related characteristics, pleural effusion (6.7% vs.
3.8%, P = 0.02), vitamin D deficiency (36.4% vs. 27.6%, P = 0.04), and anaemia (30.8% vs. 20.3%, P = 0.002) were more frequent in SLE patients, whereas thrombocytopenia was more common in pSS patients (14.7% vs.
20.8%, P = 0.003).Table 1Comparison of clinical characteristics and risk factors between patients with SLE and pSS.SLE (n = 955)pSS (n = 558)P valueDemographic characteristicsFemale sex, n (%)869 (91.0)521 (93.4)0.13Age at diagnosis, median (IQR), years33 (26-40)44 (34-57)0.005Age at thrombosis, median (IQR), years45 (32-58)58 (47-68)0.003Traditional cardiovascular risk factors, n (%)Raised BMI (>25 kg/m2)207 (21.7)115 (20.6)0.67Ever smoker102 (10.7)37 (6.6)0.01Hypertension247 (25.9)117 (21.0)0.032Diabetes39 (4.1)42 (7.5)0.006Hyperlipidaemia130 (13.6)38 (6.8)0.001Disease-related characteristics and treatment history, n (%)Nervous system involvement21 (2.2)13 (2.3)0.86Pleural effusion64 (6.7)21 (3.8)0.02Arthritis17 (1.8)17 (3.0)0.25Interstitial pneumonia67 (7.0)55 (9.9)0.16Vitamin D deficiency348 (36.4)154 (27.6)0.04Anaemia294 (30.8)113 (20.3)0.002Thrombocytopenia140 (14.7)116 (20.8)0.003Use of hydroxychloroquine ever943 (98.7)523 (93.7)0.194Laboratory markers, n (%)Low complement ever820 (85.9)327 (58.6)0.002Elevated platelet count ever33 (3.5)13 (2.3)0.28LA positivity ever172 (18.0)84 (15.1)0.139aβ2-GP1 positivity ever109 (11.4)55 (9.9)0.39aCL positivity ever97 (10.2)40 (7.2)0.07Thrombosis outcome, n (%)Overall thrombosis180 (18.8)103 (18.5)0.91VTE51 (5.3)36 (6.5)0.115AT126 (13.2)62 (11.1)0.252Abbreviation: SLE, Systemic Lupus Erythematosus. pSS, Primary Sjögren's Syndrome. IQR, interquartile range. BMI, body mass index.
LA, lupus anticoagulant. aβ2-GP1, anti-β2-glycoprotein I. aCLs, anti-cardiolipin antibodies. VTE, Venous thromboembolism. AT, arterial thromboembolism.
Data are presented as n (%) or median (IQR). P values were calculated using the chi-square test for categorical variables and the Mann-Whitney U test for continuous variables. Statistically significant P values (P < 0.05) are shown in bold.
The incidence of thrombosis in patients with SLE and pSS
According to Table 1, the cumulative incidence of overall thrombosis did not differ significantly between SLE and pSS (18.8% vs. 18.5%, P = 0.91), and the distribution of arterial (13.2% vs. 11.1%, P = 0.252) and venous events (5.3% vs.
6.5%, P = 0.115) was similar between the two groups.
The clinical characteristics of 955 SLE patients and 558 pSS patients in the thrombotic and non-thrombotic groups were then analyzed. According to Table S1, among all thrombotic events in SLE, arterial thrombosis was more common than venous thrombosis (70.0% vs. 28.3%).
Three additional patients (1.7%) experienced two distinct types of thrombotic events. Among arterial thrombosis cases, cerebral infarction was more common, accounting for 48.9% (88/180). Among venous thrombosis cases, deep vein thrombosis was more common, accounting for 23.3% (42/180) of cases.
The median age when the thrombotic event occurred was 45 years (IQR 32-58), while the median interval from diagnosis to the event was 60 months (IQR 14-96). The survival curves are presented in Fig. 1A.Fig.
1Survival curve for thrombosis in our study population. A. Survival curve for thrombosis in SLE patients.
B. Survival curve for thrombosis in pSS patients.Abbreviation: SLE, Systemic Lupus Erythematosus. pSS, Primary Sjögren's Syndrome.
For pSS patients, the incidence of thrombosis was 18.5% (103/558), with arterial thrombosis being more prevalent than venous thrombosis (60.2% vs. 35.0%). Five additional patients (4.8%) experienced two distinct types of thrombotic events.
Among the arterial thrombosis patients, 47.6% (49/103) had cerebral infarction. Among venous thrombosis cases, deep vein thrombosis was more common, accounting for 30.1% (31/103) of cases. The median age when the thrombotic event occurred was 58 years (IQR 47-68), while the median duration from the diagnosis to the occurrence of the event was 48 months (IQR 21-108).
The survival curves are presented in Fig. 1B.
Risk factors for thrombosis in patients with SLE and pSS
SLE patients
A univariate analysis revealed several variables that were significantly linked to the risk of thrombosis (P < 0.15): age at diagnosis (HR:2.529, 95% CI:1.724-3.709, P < 0.0001), history of lupus nephritis (HR:1.534, 95% CI:1.134-2.076, P = 0.006), smoking history (HR:3.791, 95% CI:2.719-5.284, P < 0.0001), hypertension (HR:1.830, 95% CI:1.352-2.476, P < 0.0001), low complement levels (HR:1.322, 95% CI:0.901-1.941, P = 0.144), positive LA (HR:3.974, 95% CI:2.944-5.363, P < 0.0001), aβ2-GP1 positivity (HR:1.583, 95% CI:1.034-2.422, P = 0.047), and aCLs positivity (HR:2.356, 95% CI:1.752-3.681, P = 0.002). Variables with P < 0.15 in univariate analysis were entered into a multivariable Cox regression model, which showed that hypertension (HR:1.523, 95% CI:1.213-2.135, P = 0.03), history of lupus nephritis (HR:1.822, 95% CI:1.225-2.883, P < 0.0001), smoking history (HR:3.613, 95% CI:2.426-4.631, P < 0.0001), and positive LA (HR:3.213, 95% CI:2.311-4.157, P < 0.0001) were independent risk factors for thrombotic events (Table 2).Table 2Univariable and multivariable Cox regression analysis for thrombosis in SLE patients.CovariateUnivariable analysisMultivariable analysisPHR (95% CI)PHR (95% CI)Female sex0.9741.000(0.980-1.020)--Age at diagnosis<0.00012.529(1.724-3.709)0.311.021(0.832-1.134)Raised BMI (>25 kg/m2)0.9401.013(0.713-1.440)--Hypertension<0.00011.830(1.352-2.476)0.031.523(1.213-2.135)Diabetes0.2791.404(0.760-2.593)--Hyperlipidaemia0.3971.184(0.801-1.749)--Lupus nephritis ever0.0061.534(1.134-2.076)<0.00011.822(1.225-2.883)Ever smoker<0.00013.791(2.719-5.284)<0.00013.613(2.426-4.631)Use of hydroxychloroquine ever0.4990.674(0.215-2.111)--Low complement ever0.1441.322(0.901-1.941)0.4520.776(0.231-1.436)Anti-dsDNA positivity ever0.8141.036(0.773-1.389)--PLT positivity ever0.2491.956(0.625-6.126)--D-dimer positivity ever0.2861.174(0.875-1.575)--LA positivity ever<0.00013.974(2.944-5.363)<0.00013.213(2.311-4.157)aβ2-GP1 positivity ever0.0471.583(1.034-2.422)0.9811.017(0.711-1.623)aCLs positivity ever0.0022.356(1.752-3.681)0.1131.042(0.814-2.436)Abbreviation: SLE, Systemic Lupus Erythematosus. HR, Hazard Ratio.
CI, confidence interval. BMI, body mass index. dsDNA, double-stranded DNA. PLT, platelet.
LA, lupus anticoagulant. aβ2-GP1, anti-β2-glycoprotein I. aCLs, anti-cardiolipin antibodies.Statistically significant P values (P < 0.05) are marked in bold in the multivariable analysis.
pSS patients
A univariate analysis showed multiple variables significantly associated with thrombosis risk (P < 0.15): age at diagnosis (HR:1.022, 95% CI:1.008-1.036, P = 0.002), elevated platelet count (HR:3.575, 95% CI:1.660-7.701, P = 0.001), positive D-dimer (HR:1.543, 95% CI:1.048-2.273, P = 0.028), positive LA (HR:2.973, 95% CI:1.966-4.495, P < 0.0001), aβ2-GP1 positivity (HR:1.227, 95% CI:0.712-2.411, P = 0.141), and aCLs positivity (HR:1.637, 95% CI:1.127-2.914, P = 0.048). Among traditional risk factors, hypertension (HR:2.397, 95% CI:1.617-3.555, P < 0.0001) and diabetes (HR:2.836, 95% CI:1.706-4.713, P < 0.0001) were significantly associated with thrombosis risk. Multivariable Cox regression incorporating variables with P < 0.15 in univariate analysis identified diabetes (HR:1.913, 95% CI:1.112-3.245, P = 0.023), elevated platelet count (HR:3.115, 95% CI:1.672-6.779, P = 0.006), and positive LA (HR:2.611, 95% CI:1.697-3.782, P < 0.0001) as independent risk factors for thrombotic events in pSS (Table 3).Table 3Univariable and multivariable Cox regression analysis for thrombosis in pSS patients.CovariateUnivariable analysisMultivariable analysisPHR (95% CI)PHR (95% CI)Female sex0.3410.732(0.370-1.450)--Age at diagnosis0.0021.022(1.008-1.036)0.1121.021(0.927-1.235)Raised BMI (>25 kg/m2)0.3381.245(0.795-1.952)--Hypertension<0.00012.397(1.617-3.555)0.1041.472(0.828-2.133)Diabetes<0.00012.836(1.706-4.713)0.0231.913(1.112-3.245)Hyperlipidaemia0.5290.767(0.336-1.750)--Ever smoker0.3711.366(0.690-2.706)--Use of hydroxychloroquine ever0.2392.028(1.133-2.305)--Low complement ever0.6920.925(0.628-1.361)--PLT positivity ever0.0013.575(1.660-7.701)0.0063.115(1.672-6.779)D-dimer positivity ever0.0281.543(1.048-2.273)0.4681.116(0.611-1.852)LA positivity ever<0.00012.973(1.966-4.495)<0.00012.611(1.697-3.782)aβ2-GP1 positivity ever0.1411.227(0.712-2.411)0.7131.135(0.598-2.315)aCLs positivity ever0.0481.637(1.127-2.914)0.6821.255(0.762-2.733)Abbreviation: pSS, Primary Sjögren's Syndrome.
HR, Hazard Ratio. CI, confidence interval. BMI, body mass index.
PLT, platelet. LA, lupus anticoagulant. aβ2-GP1, anti-β2-glycoprotein I. aCLs, anti-cardiolipin antibodies.Statistically significant P values (P < 0.05) are marked in bold in the multivariable analysis.
Risk factors for specific thrombosis types in SLE and pSS patients
SLE patients
Multivariable analysis identified the following variables as independently associated with venous thrombosis (P < 0.05): history of lupus nephritis (HR:2.151, 95% CI:1.167-3.815, P = 0.012), smoking history (HR:3.233, 95% CI:1.835-6.211, P = 0.002), and positive LA (HR:3.394, 95% CI:1.623-6.197, P < 0.0001) (Fig. 2A, Table S2). For arterial thrombosis, hypertension (HR:1.324, 95% CI:0.822-2.286, P = 0.044), history of lupus nephritis (HR:1.821, 95% CI:1.092-2.613, P = 0.007), smoking history (HR:3.365, 95% CI:2.126-5.443, P < 0.0001), and positive LA (HR:3.123, 95% CI:2.153-4.995, P < 0.0001) emerged as significant risk factors (Fig.
2B–Table S2).Fig. 2Multivariable Cox regression analysis to evaluate the association of independent variables with different types of thrombosis in the overall study cohort. A.
Multivariable Cox regression analysis to evaluate the association of independent variables with venous thrombosis in SLE patients. B. Multivariable Cox regression analysis to evaluate the association of independent variables with arterial thrombosis in SLE patients.
C. Multivariable Cox regression analysis to evaluate the association of independent variables with venous thrombosis in pSS patients. D.
Multivariable Cox regression analysis to evaluate the association of independent variables with arterial thrombosis in pSS patients.Abbreviation: SLE, Systemic Lupus Erythematosus. HR, Hazard Ratio. CI, confidence interval.
LA, lupus anticoagulant. aβ2-GP1, anti-β2-glycoprotein I. aCLs, anti-cardiolipin antibodies.Statistically significant P values (P < 0.05) are marked in bold in the multivariable analysis.
pSS patients
In the pSS cohort, positive LA was identified as the only independent risk factor associated with venous thrombosis (HR:4.683, 95% CI:2.435-7.345, P < 0.0001) (Fig. 2C–Table S3). Independent predictors of arterial thrombosis included hypertension (HR:1.916, 95% CI:1.013-3.433, P = 0.03), diabetes (HR:2.523, 95% CI:1.246-4.822, P = 0.014), elevated platelet count (HR:3.975, 95% CI:1.423-8.232, P = 0.012), and positive LA (HR:1.965, 95% CI:1.042-3.881, P = 0.041) (Fig.
2D–Table S3).
Creation and assessment of thrombosis risk models
To further quantify the individualized risk of thrombotic events in SLE and pSS patients, we built a prediction model for SLE using the independent risk factors derived from multivariable analysis (hypertension, history of lupus nephritis, smoking history, and positive LA). For pSS, we assessed the combined discriminative performance of diabetes, elevated platelet count, and positive LA.
With the date of diagnosis of the autoimmune disease as the starting point and the date of the first thrombotic event as the endpoint, we assessed the discriminative ability of each model at 3, 5, and 10 years by employing time-dependent ROC curves, while calculating the AUCs. To avoid overfitting and obtain a more robust estimate of generalizability, we performed internal validation using bootstrap resampling (1000 iterations) to obtain bias-corrected AUCs. Additionally, Harrell's C-index was computed to globally evaluate the ability to rank patients by thrombotic risk.
For the SLE prediction model, the AUCs at 3, 5, and 10 years were 0.778, 0.761, and 0.755, respectively, and the C-index was 0.752 (95% CI:0.71-0.79) (Fig. 3A–D). In pSS, the combined discriminative capacity of the risk factors was limited, with AUCs of 0.600, 0.606, and 0.639, and a C-index of 0.665 (95% CI:0.587-0.743) (Fig.
3E–H). The pSS performance showed inferior discrimination compared with the SLE model, particularly for short-term prediction. When each predictor was assessed individually in pSS, positive LA alone yielded a C-index of 0.612, elevated platelet count 0.574, and diabetes 0.568.
NRI analysis revealed that adding any single predictor did not significantly improve risk reclassification (all P > 0.05). Thus, these routine variables alone cannot reliably identify high-risk individuals in pSS.Fig. 3Evaluation of thrombosis risk discrimination in SLE and pSS.
A. Time-dependent ROC curves for predicting 3-, 5-, and 10-year overall thrombosis events in SLE patients. B.
The ROC analysis of 3, 5, and 10 years with VTE in SLE patients. C. The ROC analysis of 3, 5, and 10 years with AT in SLE patients.
D. Harrell's C-index of the prediction model in SLE patients. E.
Time-dependent ROC curves for predicting 3-, 5-, and 10-year overall thrombosis events in pSS patients. F. The ROC analysis of 3, 5, and 10 years with VTE in pSS patients.
G. The ROC analysis of 3, 5, and 10 years with AT in pSS patients. H.
Harrell's C-index of the identified risk factors in pSS patients.Abbreviations: SLE, Systemic Lupus Erythematosus. pSS, Primary Sjögren's Syndrome. AVT, Arterial and venous thromboembolism. VTE, venous thromboembolism.
AT, arterial thromboembolism.
We further compared the performance across thrombosis subtypes. For the SLE prediction model, the AUCs for venous thrombosis at 3, 5, and 10 years were 0.797, 0.762, and 0.736, respectively, and the C-index was 0.746 (95% CI:0.676-0.805) (Fig. 3B–D); the AUCs for arterial thrombosis were 0.778, 0.768, and 0.756, with a C-index of 0.746 (95% CI:0.709-0.782) (Fig.
3C and D). For pSS, the combined risk factors for venous thrombosis yielded AUCs of 0.589, 0.619, and 0.663, and the C-index was 0.626 (95% CI:0.576-0.677) (Fig. 3F–H); for arterial thrombosis, the AUCs were 0.677, 0.679, and 0.699, with a C-index of 0.685 (95% CI:0.617-0.754) (Fig.
3G and H).
Likelihood ratios at the Youden-optimal cutoffs are summarized in Table 4. For the SLE model, LR+ and LR− at 3, 5, and 10 years were 3.23 (95% CI 2.17-3.94) and 0.39 (0.21-0.49), 2.56 (2.20-3.71) and 0.35 (0.25-0.49), and 3.07 (2.28-4.34) and 0.42 (0.31-0.55), respectively, indicating that classification above the optimal cutoff approximately tripled the odds of a thrombotic event. For the pSS cohort, LR + values were only 1.84 (1.35-5.04), 1.88 (1.49-6.75), and 2.32 (1.41-8.29) at 3, 5, and 10 years, with corresponding LR− of 0.77 (0.58-0.93), 0.77 (0.59-0.91), and 0.71 (0.53-0.91).
These likelihood ratios close to unity further indicate that the combined factors could not reliably identify high-risk individuals with pSS.Table 4Likelihood ratios at the Youden-optimal cutoffs of the risk score for thrombotic events in patients with SLE and pSS.CohortTime (years)SensitivitySpecificityLR+ (95% CI)LR- (95% CI)SLE (AVT)30.6930.7853.23 (2.17-3.94)0.391 (0.213-0.488)SLE (AVT)50.7510.7072.56 (2.20-3.71)0.353 (0.246-0.493)SLE (AVT)100.6700.7823.07 (2.28-4.34)0.422 (0.310-0.548)SLE (VTE)30.6960.8083.63 (1.90-6.04)0.377 (0.056-0.551)SLE (VTE)50.6520.8113.46 (1.82-7.14)0.429 (0.180-0.632)SLE (VTE)100.7220.6462.04 (1.67-9.36)0.431 (0.200-0.725)SLE (AT)30.6920.7933.34 (2.19-4.34)0.388 (0.225-0.519)SLE (AT)50.7500.7172.65 (2.15-3.95)0.349 (0.224-0.498)SLE (AT)100.6720.7843.11 (2.34-5.15)0.419 (0.291-0.568)pSS (AVT)30.3990.7841.84 (1.35-5.04)0.767 (0.579-0.934)pSS (AVT)50.3960.7891.88 (1.49-6.75)0.765 (0.592-0.905)pSS (AVT)100.4160.8212.32 (1.41-8.29)0.711 (0.528-0.911)pSS (VTE)30.3120.8662.32 (1.14-4.50)0.795 (0.503-1.000)pSS (VTE)50.3780.8642.77 (1.25-4.73)0.720 (0.461-0.980)pSS (VTE)100.3050.9213.86 (1.29-14.84)0.755 (0.549-0.964)pSS (AT)30.6200.6781.93 (1.54-7.05)0.560 (0.314-0.807)pSS (AT)50.5510.7852.56 (1.64-7.90)0.573 (0.346-0.769)pSS (AT)100.6780.7442.64 (1.59-6.55)0.434 (0.218-0.715)Abbreviations: SLE, Systemic Lupus Erythematosus. pSS, Primary Sjögren's Syndrome. AVT, Arterial and venous thromboembolism. VTE, venous thromboembolism.
AT, arterial thromboembolism. LR+, positive likelihood ratio. LR−, negative likelihood ratio.
CI, confidence interval. 95% CIs were estimated from 1000 bootstrap resamples.
Discussion
In this large, retrospective cohort study, we systematically characterized risk factor profiles for overall thrombotic events, as well as for arterial and venous thrombosis subtypes, in Chinese patients diagnosed with SLE and pSS, and constructed corresponding risk prediction models. The main results are summarized below. First, the cumulative incidence of thrombotic events was similar among SLE and pSS (18.8% vs 18.5%).
Arterial events dominated over venous events in both diseases (SLE: 70.0% vs 28.3%; pSS: 60.2% vs 35.0%), with cerebral infarction and deep vein thrombosis being the most frequent arterial and venous manifestations, respectively. Second, the independent risk factors associated with overall thrombosis in individuals with SLE included history of lupus nephritis, hypertension, smoking and positive LA. For pSS, those were diabetes, elevated platelet count and positive LA.
Third, in SLE, positive LA, smoking, and history of lupus nephritis were associated with both arterial and venous thrombosis, hypertension was exclusively associated with arterial events. In pSS, positive LA was the only independent predictor of venous thrombosis, while independent predictors of arterial thrombosis were hypertension, diabetes, elevated platelet count and positive LA. Furthermore, the SLE prediction model showed acceptable-good discriminative performance (C-index, 0.752), whereas the combination of risk factors in pSS had limited discriminative ability (C-index, 0.665).
While many previous studies have looked at risk factors for thrombosis in autoimmune diseases, these studies have been fairly heterogeneous. Table 5 depicts the findings of studies with larger scales conducted recently [18], [19], [20], [21], [22], [23], [24], [25], [26], [27], [28], [29], [30], [31], [32], [33], [34], [35], [36]. There are significant inconsistencies in reported thrombotic events’ prevalence and potential predictors.
This variability may be due to ethnicity, geographic location and number of patients and centers taking part in a trial, and duration of follow-up. Unlike previous studies, the present study is the first to systematically compare risk-factor profiles for arterial and venous thrombosis between SLE and pSS in a large Asian cohort over a 15-year follow-up.Table 5Over the past ten years, retrospective and prospective cohort studies on thrombotic events in autoimmune diseases are summarized in the table below.Authors and yearStudy designAutoimmune disease typePatients numberRegionFollow up durationTypes of thrombosisFactors associated with thrombosisPMIDGalarza D et al., 2026Retrospective cohort studyAPS211EuropeUp to 10 yearsVTEthrombocytopenia41297380Kharouf F et al., 2026Prospective cohort studyLN324North AmericaUp to 9.9 yearsVTEthe SLEDAI-2K score, proteinuria level40891890Murray A et al., 2024Retrospective cohort studySLE341North AmericaUp to 15 yearsVTE, ATIgG aPS/PT39672566Mittal P et al., 2024Retrospective cohort studySLE118EuropeUp to 42 yearsVTE, ATLA (lupus anticoagulant)39092568Santos CS et al., 2023Retrospective cohort studySS102EuropeUp to 20 yearsATAnti-Ro/SSA and anti-La/SSB37402929Nicola Farina et al., 2023Retrospective cohort studySLE419EuropeUp to 40 yearsVTE, ATantiphospholipid antibodies37201214Rong X et al., 2022Retrospective cohort studyITP1824AsiaUp to 10 yearsATAge36054971Park DJ et al., 2021Prospective cohort studySLE259AsiaUp to 103 monthsVTE, AThypertension, higher mean daily dose of prednisolone, diagnosis of APS, and higher SDI41468843Haugaard et al., 2021Retrospective cohort studySLE4587EuropeUp to 20 yearsATNon-use of HCQ33321159Liapi M et al., 2021Retrospective cohort studyAAV325EuropeUp to 20 yearsVTEage and Birmingham Vasculitis Activity Score (BVAS)33506869Zuily S et al., 2020Prospective cohort studySLE/APS137EuropeMean 43.1 monthsVTE, ATAnti-domain I β2-GPI antibodies (aDI), activated protein C resistance32572467Ito S et al., 2020Retrospective cohort studyITP303AsiaUp to 16.7 yearsVTE, ATSmoking, atrial fibrillation, male gender, history of thrombosis31853702Desai RJ et al., 2020Retrospective cohort studyRA50865North AmericaNot reportedVTETofacitinib30552833Isaacs B et al., 2020Retrospective cohort studyAAV162North AmericaMedian time to VTE 1 monthVTEPR3-ANCA positivity, hypoalbuminemia, high BMI35372925Urowitz et al., 2020Prospective cohort studySLE1710America/Europe/AsiaUp to 18 yearsATBMI >40 kg/m2, prior CVE, non-use of HCQ32515554Yoo J et al., 2019Retrospective cohort studyAAV138AsiaMean 50.8 monthsVTE, ATPersistent antiphospholipid antibody positivity, age at diagnosis30798998Ekstrand C et al., 2019Retrospective cohort studyITP9715EuropeFrance, 2009-2015. Sweden, 2009-2016VTE, ATOlder age, male sex, prior AT. older age, prior VTE, cancer31026663Mofors J et al., 2019Retrospective cohort studypSS960EuropeMedian 9.5 yearsVTE, ATRo/SSA and La/SSB double positivity31127862Kang A et al., 2018Retrospective cohort studyAAV204EuropeMedian 5.8 yearsVTE, ATOlder age, prior IHD. active disease, hospitalization30385704Demetrio Pablo R et al., 2017Retrospective cohort studyaPL carriers138EuropeMean 12.2 yearsVTE, ATThrombocytopenia28153433Avina-Zubieta JA et al., 2017Retrospective cohort studypSS1175North AmericaUp to 14 yearsVTEpSS28298559Ly KH et al., 2016Retrospective cohort studyGCA428EuropeMean 61.2 monthsVTEPrior venous thrombosis, varicose veins, higher glucocorticoid dose27324606Unizony S et al., 2016Retrospective cohort studyGCA6414EuropeMedian 5.2 yearsVTEGCA28029234Habe K et al., 2016Retrospective cohort studySLE/SSc/SS/MCTD/ITP233AsiaUp to 19 yearsVTE, ATLupus anticoagulant, aCL IgG, anti-β2-GPI complex IgG, antiphospholipid antibody positivity26984073Hinojosa-Azaola A et al., 2016Retrospective cohort studySLE219North AmericaMedian 5.21 yearsVTE, ATCutaneous vasculitis, nephrotic syndrome, prednisone dose, LAC + anti-RNP/Sm, older age, traditional risk factors, cumulative prednisone dose26773122Bleau N et al., 2016Retrospective cohort studySLE/DM/RA/T1D/IBD43,523North AmericaUp to 8 yearsVTESLE, dermatomyositis, RA, type 1 diabetes, ulcerative colitis, Crohn's disease25294635Abbreviation: AAV, ANCA-associated vasculitis. aPL, antiphospholipid antibody.
APS, antiphospholipid syndrome. DM, dermatomyositis. GCA, giant cell arteritis.
IBD, inflammatory bowel disease. ITP, immune thrombocytopenia. LN, lupus nephritis.
MCTD, mixed connective tissue disease. pSS, primary Sjögren's syndrome. RA, rheumatoid arthritis. SLE, systemic lupus erythematosus.
SS, Sjögren's syndrome. SSc, systemic sclerosis. T1D, type 1 diabetes.
The presence of LA is acknowledged as a risk factor for unexplained recurrent miscarriages, fetal demise, fetal growth restriction, venous and arterial embolism, thrombophilia, and specific autoimmune diseases [37], [38]. Our findings identify LA as the only risk factor that consistently demonstrates an independent association with different types of thrombotic events across both major categories of autoimmune diseases. Among SLE patients, those who were LA-positive had an approximately threefold higher risk of thrombosis compared with LA-negative patients (HR:3.213), and LA positivity was associated with both arterial and venous events (HR 3.123 and 3.394).
In the pSS cohort, a positive result for LA was associated with approximately a 2.6-fold heightened risk of thrombosis (HR: 2.611) and was identified as the only independent predictor of venous thrombosis, demonstrating a hazard ratio reaching up to 4.683. LAs, along with aβ2-GP1 and aCLs, are collectively known as aPLs. The downstream effects of their interaction with vascular endothelial cells can increase the potential for local thrombosis; their activation of platelets increases the risk of thrombosis, and their influence on monocytes induces the release of inflammatory factors and the expression of tissue factor.
In addition, they can affect neutrophils, trigger complement activation, and provoke immune responses through complement pathways, all of which may promote thrombosis [39], [40], [41]. It should be noted that while aβ2-GP1 and aCLs were significant in univariate analysis, neither retained an independent effect in the multivariable model. These findings indicate that, among the antiphospholipid antibody spectrum, LA is the strongest independent correlate of thrombotic complications in both autoimmune diseases.
Patients with SLE and pSS who are LA-positive should be considered at high risk of thromboembolism during routine clinical follow-up.
SLE and pSS presented different distribution patterns of traditional thrombosis risk factors. Previous studies have indicated that the prevalence of smoking, dyslipidaemia, and hypertension is lower in patients with pSS compared to those with SLE [14], which aligns with the findings of our study. Within the SLE cohort, a history of smoking was a particularly strong independent risk factor (overall HR:3.613).
The use of tobacco increases oxidative stress; endothelial dysfunction; platelet activation; increases procoagulant factors; reduces activity of anticoagulant proteins; thus, results in a hypercoagulable state. This effect compounds the immune dysregulation inherent to lupus, further promoting thrombus formation [38], [42]. Thromboprophylaxis in SLE patients should include smoking cessation as an essential measure.
Our findings further indicate that hypertension is associated with arterial thrombotic risk in both SLE and pSS. In patients with hypertension, there may be endothelial cell damage, abnormal blood flow, and blood component abnormalities, including coagulation or hemostatic factors and platelet activation, that lead to a pro-thrombotic state [43]. Previous studies have shown that diabetes serves as an independent risk factor for thrombosis, the pro-thrombotic state of diabetic patients results from the effectiveness of various mechanisms, such as endothelial dysfunction, activation of coagulation and increased platelet reactivity [44].
In contrast, diabetes does not act as an independent risk factor for thrombosis in SLE. This is probably because its association with thrombotic risk is hidden by other stronger associated drivers such as LA positivity, lupus nephritis and smoking. This variation indicates that metabolic factors may play a more independent role in the onset of thrombosis in pSS.
This research found a significant association between lupus nephritis and both arterial and venous events in patients with SLE. Earlier studies showed a link between lupus nephritis and thrombotic phenomena. Chronic inflammation and the accumulation of immune complexes in the renal tissue of individuals with lupus nephritis led to increased vascular damage, contributing to a hypercoagulable state [10], [45].Thus, SLE patients with kidney involvement should be closely monitored for thrombotic risk.
A salient observation in the pSS cohort is that high platelet count was a strong independent predictor of arterial thrombosis (HR:3.975). An elevated platelet count in pSS may reflect ongoing systemic inflammation. Beyond increasing blood viscosity, activated platelets release microparticles, adhesion molecules, and tissue factor, thereby driving immunothrombosis [46], [47].
Considering that platelet count is such a routine and easy-to-access laboratory parameter, it deserves specific attention to its value in thrombotic risk assessment in pSS.
The discriminative performance of these risk factor combinations underwent internal validation and showed notable differences between the two diseases. The SLE prediction model exhibited satisfactory discrimination (C-index 0.752; 3-year AUC 0.778). In contrast, the combined pSS risk factors showed poor discrimination (C-index 0.665; 3-year AUC 0.600).
Stratification by thrombosis subtype revealed that the SLE venous model provided the best discrimination (3-year AUC 0.797), whereas the pSS venous model performed the worst (3-year AUC 0.589). The limited number of venous thrombosis events in the pSS cohort (n = 36) may partially account for these discrepancies, thereby constraining predictive performance. Conversely, the SLE model benefited from the inclusion of well-characterized and robust predictors, specifically LA, lupus nephritis, and smoking history.
The limited discriminative performance of the pSS predictors indicates that these routine clinical variables alone cannot support clinical decision-making for thromboprophylaxis in pSS. The independent associations of LA, diabetes, and elevated platelet count with thrombosis remain etiologically informative, but their translation into risk stratification will require integration with deeper molecular phenotyping, such as proteomic or metabolomic profiling, in future prospective studies.
In our multivariable model, hydroxychloroquine use was not associated with a decreased risk of thrombosis in either SLE or pSS patients. However, hydroxychloroquine was used by the vast majority of patients in both the thrombosis and non-thrombosis groups, which precluded a meaningful comparison between users and non-users. Therefore, no conclusions regarding its antithrombotic effect can be drawn from this study.
Previous studies have indicated that hydroxychloroquine exerts antithrombotic effects by inhibiting platelet activation, downregulating tissue factor expression, and attenuating the procoagulant activity of antiphospholipid antibodies [48], [49], while others have suggested it may be insufficient to prevent platelet-mediated thrombosis [50]. This question requires evaluation in cohorts with more heterogeneous prescription patterns.
This research has several limitations. Firstly, it utilizes a retrospective design based in a single center, and the lack of validation from external sources undermines the reliability of the established model. Therefore, future multicenter cohort studies are necessary to further validate the current findings.
Secondly, our model did not incorporate disease-related variables that changed over time, especially disease activity (such as the SLE Disease Activity Index [SLEDAI]) and cumulative organ damage (such as the SLICC/ACR Damage Index [SDI]). Increasing evidence, including a comprehensive review recently published by Shui et al. [22], indicates that both disease activity and cumulative damage are closely associated with the risk of thrombosis in SLE. High disease activity drives pro-inflammatory and pro-thrombotic states through immune complex deposition, complement activation, endothelial dysfunction and elevated reactants in the acute phase.
Chronic organ damage, on the other hand, reflects long-term vascular damage and is particularly associated with arterial thrombotic events. Because these variables evolve dynamically during the disease process, evaluations based solely on baselines or fixed time points may underestimate their true impact on longitudinal risk. It is necessary to conduct prospective studies in the future, including a series of measurements of disease activity and damage as time-varying covariates, in order to further optimize the thrombosis risk stratification.
Additionally, given the retrospective design, the identified associations should not be interpreted as causal relationships. Furthermore, the study did not systematically collect data on concomitant medications that could influence thrombotic risk, such as anticoagulants, antiplatelet agents, and glucocorticoid dosages, which may have confounded the results.
Conclusion
In conclusion, this large long-term retrospective cohort study systematically compared thrombotic incidence and risk profiles between Chinese SLE and pSS patients based on updated 2023 ACR/EULAR aPL diagnostic criteria. LA positivity was the sole consistent independent risk factor for both arterial and venous thrombosis across SLE and pSS populations. In SLE patients, the prediction model comprising LA positivity, hypertension, smoking history, and lupus nephritis showed acceptable-good discriminative ability for thrombotic risk stratification.
In pSS, LA positivity, hypertension, diabetes, and elevated platelet count were independently associated with thrombosis, but their combined predictive performance was limited. Clinically, targeted interventions for SLE patients should focus on smoking cessation, standardized management of lupus nephritis and regular systematic monitoring of LA. For pSS, attention to LA, hypertension, diabetes, and platelet count remains reasonable in comprehensive disease management, but formal risk stratification will require more robust predictive tools.
CRediT authorship contribution statement
Mao Xia: Writing – review & editing. Daibao Peng: Investigation. Hui Xie: Validation.
Guanghao Song: Methodology. Guiting Zhang: Formal analysis. Sijia He: Investigation.
Haixuan Sun: Project administration. Zhiming Ding: Investigation. Han Shen: Formal analysis.
Fei Chen: Writing – original draft, Data curation.
Ethics declaration
Written informed consent to take part in the study and to publish the article has been obtained from all participants or their legal representatives. The privacy rights of participants have been observed.
This study was performed in compliance with relevant laws, regulatory frameworks and guidelines where the research took place. This study was approved by the Ethics Committee of Nanjing Drum Tower Hospital and Nanjing University Medical School. (Approval No.
2022-360-03)
Ethics approval and consent to participate
This study followed the Declaration of Helsinki. The Ethics Committee of Nanjing Drum Tower Hospital and Nanjing University Medical School approved (Project Number: 2022-360-03) the current study. Acknowledging the study's retrospective nature, the ethics committee permitted waiver of informed consent.
All data have been depersonalized.
Disclosure statement
No potential conflict of interest was reported by the authors.
Consent for publication
Not applicable.
Funding
This study was supported by grants from Jiangsu Provincial Research Hospital (YJXYY202204-2-ZD03 to M.X.), 10.13039/501100012166National Key Research and Development Program of China (2023YFC2415000 to M.X.), Nanjing Medical Science and Technology Development Foundation (YKK23094 to M.X.), Mandatory Project of Nantong Municipal Health Commission (MS2022094 to Z.M.D), Scientific Research Project of Taikang Xianlin Drum Tower Hospital, the Affiliated Hospital of Nanjing University Medical School (TKKYYB20211607 to Daibao Peng), 10.13039/501100001809National Natural Science Foundation of China (82402114 to Guiting Zhang), 10.13039/501100004608Natural Science Foundation of Jiangsu Province (BK20240237 to Guiting Zhang) and Clinical Trials from the Affiliated Drum Tower Hospital, Medical School of 10.13039/501100008048Nanjing University (2022-LCYJ-PY-20 to M.X.).
Declaration of competing interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.