Abstract
Objective We aimed to evaluate the performance of cardiovascular (CV) risk prediction tools in identifying patients with systemic lupus erythematosus (SLE) with coronary artery calcification (CAC).
Methods We conducted a post hoc analysis of a prior case-control study of adult female patients with SLE and matched controls. We excluded patients who already met cardiology guidelines for recommended statin use. We risk-stratified patients per Atherosclerotic CV Disease (ASCVD) Risk Score and SLE-specific CV Risk (SLECRISK) score and determined rates of abnormal CAC. We used 2-sample t tests, 2-sample Wilcoxon rank-sum (Mann-Whitney) tests, and chi-square tests to compare various characteristics between subgroups, and used logistic regression to assess adjusted risk for patients with SLE compared to controls.
Results We analyzed 128 patients with SLE and 138 controls. Rates of abnormal CAC in the SLE and control groups were 31.3% (40/128) and 12.3% (17/138), respectively, with similar mean ASCVD Risk Scores (2.2% vs 1.9%). Both the ASCVD Risk Score and SLECRISK score had relatively high specificity (95-100%) for abnormal CAC. The ASCVD score demonstrated poor sensitivity in both control (17.6%) and SLE (15%) groups, with sensitivity doubling to 30.8% among patients with SLE with the use of the SLECRISK score. Among participants with SLE who had low ASCVD risk, those with abnormal CAC were more likely to be older and have lower glomerular filtration rate, longer disease duration, higher insulin resistance, and higher low-density lipoprotein and cholesterol levels.
Conclusion The poor performance of conventional and even modified risk scores suggests a continued need for screening approaches, including CAC, to determine CV disease risk in the SLE patient population.
Patients with systemic lupus erythematosus (SLE) demonstrate a high rate of cardiovascular disease (CVD)-related mortality.1,2 In fact, patients with SLE have a 5- to 9-fold higher myocardial infarction (MI) risk and 2- to 3-fold higher stroke risk when compared to the general population.3,4 Patients with SLE are also predisposed to adverse cardiac events at a younger age. Female patients with SLE (age 35-44 years) had a 52-fold higher risk for MI than age and gender-matched controls within the Framingham study (risk ratio 52.43, 95% CI 21.6-98.5).5 Moreover, patients with lupus nephritis have a significantly increased risk of MI and CVD-related mortality when compared to SLE patients without lupus nephritis.6,7
The high prevalence of CVD in the SLE patient population has raised awareness about the need for more effective risk stratification and screening modalities. Conventional tools include the Atherosclerotic CVD (ASCVD) Risk Score, as developed by the American Heart Association (AHA) and the American College of Cardiology (ACC) in accordance with traditional Framingham risk factors. However, scores such as ASCVD may underestimate CVD risk in SLE due to the presence of unique risk factors.8 CVD in patients with SLE is likely driven by immunologic and inflammatory mechanisms.9,10 Namely, the type 1 interferon pathway, proinflammatory neutrophils, and oxidized high-density lipoprotein have been implicated in SLE-related atherosclerosis,11,12 in addition to organ damage such as kidney injury.13 Other SLE-specific risk factors include higher plasma levels of oxidized low-density lipoproteins (LDL), dyslipidemia, and elevated homocysteine levels.14,15
Higher risks observed in SLE have prompted recommendations to initiate statins at a lower LDL threshold of 2.6 mmol/L.16 However, SLE is a variable disease and applying such broad recommendations can result in imprecise care. This has paved the way for newer scores such as the QRISK3, which incorporates autoimmune-related risk factors. QRISK3 has been found to better predict CVD in a cohort of patients with SLE when compared to the ASCVD score and Framingham Risk Score (FRS).17 Moreover, a modified ASCVD Risk Score, the SLE-specific CV risk (SLECRISK) score, identified twice the number of SLE patients at high risk for CVD events over 10 years.18
Despite the emergence of new risk scores, many of them have been found to still underestimate CVD risk compared to assessment of atherosclerotic plaques on imaging modalities such as computed tomography (CT) and ultrasound (US).19,20 Imaging biomarkers, such as coronary artery calcification (CAC) scores, may serve as another method to improve the accuracy of CVD risk stratification in SLE. SLE has been associated with a higher presence of CAC when compared to age- and sex-matched controls (odds ratio [OR] 30.0, 95% CI 6.7-203.8).21,22 The presence of CAC is highly predictive of CVD events, and statin therapy is typically prescribed in patients with abnormal CAC scores.
These findings suggest that there may be value in incorporating CAC scores as a secondary screening measure for patients with SLE. We aimed to determine how many patients with SLE and low-risk ASCVD Risk Score and SLECRISK score have abnormal CAC scores, suggesting that they might benefit from statin therapy. We also aimed to highlight clinical factors associated with CAC in this identified low-risk group to inform clinicians which of these patients with SLE would most likely benefit from CT coronary imaging for CVD risk stratification and statin initiation.
METHODS
This is a post hoc analysis of a prior case-control study, which was approved by the University of Pennsylvania Institutional Review Board.14 No new data were collected from human subjects for our analysis. Female, nonpregnant patients (aged > 18 years) with SLE and matched controls were recruited from a single urban academic health center. All patients underwent a history and physical examination, electrocardiogram (ECG), coronary artery CT, and fasting blood sample analysis, which was used to determine lipid profile, antibody levels, blood cell counts, and inflammatory markers. SLE disease activity and organ damage were determined using the SLE Disease Activity Index (SLEDAI) and the Systemic Lupus International Collaborating Clinics/American College of Rheumatology Damage Index. Further methodological overview is available in the original study.14
We excluded patients with history of MI, cerebrovascular accident (CVA), diabetes mellitus (DM), or LDL ≥ 190 mg/dL because current cardiology guidelines recommend statin use in these patients. We risk-stratified SLE and control patients per ASCVD (ACC/AHA 2013) and SLECRISK criteria (Supplementary Table S1, available from the authors upon request). Low ASCVD risk was defined as score < 7.5%. We used the SLECRISK formula as described in the original paper,14 with a few assumptions.18 Lower limits for C4 and dsDNA were assumed to be < 15 and < 50 mg/100 mL, respectively. If the patient did not have lupus anticoagulant data available (about half of participants), we presumed a positive result if thrombin time index > 1.4 and/or partial thromboplastin time > 42 seconds. We further determined rates of CAC within the low-risk and high-risk categories. We considered normal CAC to be an Agatston score of 0, and abnormal CAC to be a score > 0. We further stratified abnormal CAC into low CAC (score 1-99) and moderate-to-high CAC (score > 100).
We used 2-sample t tests, 2-sample Wilcoxon rank-sum (Mann-Whitney) tests, and chi-square tests to compare various characteristics between low-risk controls and SLE patients with normal and abnormal CAC scores. We then used logistic regression to assess adjusted risk for patients with SLE compared to controls.
RESULTS
The original study included 165 patients with SLE and 160 age-matched and race-matched controls (Figure 1). A total of 55 participants were excluded for history of MI (n = 6), CVA (n = 24), DM (n = 23), and high LDL (n = 11), with 129 patients with SLE and 141 controls remaining. Patients with SLE were more likely to have a history of MI (n = 4), CVA (n = 20), and DM (n = 17) but less likely to have elevated LDL (n = 2). Of the patients of interest, 1 patient with SLE and 3 controls had missing data, resulting in an inability to calculate ASCVD risk. Statin data (use vs no use) were unavailable for 1 patient with SLE, and sufficient information was not present to determine SLECRISK for 2 patients.
ASCVD Risk Score risk stratification by CAC levels. Moderate-to-high CAC: > 100 Agatston score; low CAC: 1-99 Agatston score. * 1 SLE, 3 controls with missing data and unable to calculate ASCVD risk. † 1 SLE with missing statin data. ASCVD: atherosclerotic cardiovascular disease; CAC: coronary artery calcification; CVA: cerebrovascular accident; DM: diabetes mellitus; LDL: low-density lipoprotein; MI: myocardial infarction; mod: moderate; SLE: systemic lupus erythematosus.
The mean ASCVD Risk Score among 138 controls was 1.9%, whereas among patients with SLE, it was 2.2%. About 10.9% (14/129) of control participants with low ASCVD Risk Scores had abnormal CAC (89.1% negative predictive value [NPV]; Table 1). Rates of abnormal CAC in the SLE and control groups were 31.3% (40/128) and 12.3% (17/138), respectively. In contrast, 27.9% (34/122) of patients with SLE with low ASCVD risk had abnormal CAC (72.1% NPV). One-third (3/9) of the control participants with high ASCVD risk had an abnormal CAC score (33.3% positive predictive value [PPV]), whereas all (6/6) of the patients with SLE with high ASCVD Risk Score had abnormal CAC scores (100% PPV; Table 1).
Diagnostic performance of cardiovascular risk scores in identifying CAC in controls and participants with SLE.
Among the participants with abnormal CAC scores, 85% (34/40) of patients with SLE and 82.4% (14/17) of control participants were characterized as low ASCVD risk (15% and 17.6% sensitivity, respectively). Specificity was similar between the SLE and control groups. All patients with SLE with normal CAC were low ASCVD risk (88/88, 100% specificity) whereas only 115/121 of control participants with normal CAC were low ASCVD risk (95% specificity; Table 1). Accuracy was lower among patients with SLE compared to control participants (73.4% vs 85.5%).
Sixteen of 127 (12.6%) patients with SLE had high SLECRISK scores, whereas only 6/128 (4.7%) had high ASCVD Risk Scores (Table 1). Whereas all the 6 patients with SLE with high ASCVD Risk Scores were found to have abnormal CAC, only 12/16 (75% PPV) patients with SLE with high SLECRISK scores had abnormal CAC (4 with moderate-to-high CAC, 8 with low CAC). All patients with SLE with normal CAC had low ASCVD Risk Scores and 84/88 (95.5% specificity) had low SLECRISK scores. A smaller proportion (27/111; 24.3%) of patients with SLE with low SLECRISK scores had abnormal CAC (5 moderate-to-high CAC, 22 low CAC), resulting in a higher NPV (75.7%) compared to that of the ASCVD Risk Score (72.1%). SLECRISK captured 6 more participants as high risk, with double the sensitivity (12/39; 30.8%) compared to the ASCVD Risk Score (Figure 2).
Diagnostic characteristics of ASCVD Risk Score and SLECRISK scores for the identification of coronary artery calcification in participants with SLE. ASCVD: atherosclerotic cardiovascular disease; NPV: negative predictive value; PPV: positive predictive value; SLE: systemic lupus erythematosus; SLECRISK: SLE cardiovascular risk.
Among all participants with low ASCVD risk, patients with SLE were significantly more likely to have abnormal CAC compared with controls, adjusting for age and race (OR 3.61, 95% CI 1.76-7.41; P < 0.001). Participants with SLE with abnormal CAC were also more likely to have moderate-to-high CAC scores compared to controls, in whom all abnormal CAC scores were rated as low.
Among participants with SLE who were low ASCVD risk (mean age 41.8 years [range 19-65]), those with abnormal CAC were older (47.2 vs 39.7, P < 0.001), had lower glomerular filtration rate (GFR; 83.7 vs 98.3, P = 0.02), longer disease duration (14.3 vs 6.2, P = 0.005), higher Homeostatic Model Assessment for Insulin Resistance (HOMA-IR) scores (3.4 vs 2.2, P = 0.02), higher LDL (117.5 vs 104.2, P = 0.04), and higher total cholesterol (203.7 vs 180.5, P = 0.004; Table 2). Among those with low ASCVD Risk Scores and abnormal CAC, there was no difference between patients with SLE and control participants except for years of prednisone use and beta blocker use (Table 3).
Comparison of normal vs abnormal coronary artery calcification scores for participants with SLE with low ASCVD risk.
Comparison of SLE vs control participants with low ASCVD risk and abnormal coronary calcium scores.
Of note, all patients with SLE aged < 40 years (49/129) were identified to be low ASCVD risk, with 12.2% (6/49) demonstrating abnormal CAC on imaging. Adjusted analyses for patients with SLE aged > 40 years, the age cutoff for scores such as the ASCVD Risk Score, revealed higher sensitivity (17.7 vs 15%), similar specificity and PPV (100 vs 100%), and lower NPV (61.6 vs 72.1%; Table 4). SLECRISK analysis on patients with SLE aged > 40 years demonstrated higher sensitivity (36.4 vs 30.8%) and PPV (85.7 vs 75%), similar specificity (95.6 vs 95.5%), and lower NPV (67.2 vs 75.7%).
Diagnostic performance of cardiovascular risk scores in identifying CAC in participants with SLE aged > 40 years.
In the full cohort, 12.2% (20/164) of all patients with SLE were prescribed a statin compared with 6.9% (11/160) of controls. In contrast, only 11.8% (4/34) of patients with SLE with low ASCVD Risk Scores and abnormal CAC were taking a statin. None (0/6) of the patients with SLE with high ASCVD risk and abnormal CAC were on a statin.
DISCUSSION
Our study suggests that the higher risk of CVD in SLE is observed even among those at low calculated risk. Among patients with low ASCVD risk, rates of CAC were substantially higher for SLE (27.9%) compared to controls (10.9%). Our findings also suggest poor performance of both conventional and modified risk prediction scores for the SLE population, even in those aged > 40 years, at least in identifying those with evidence of coronary atherosclerosis.
There was high specificity for ASCVD Risk Scores and SLECRISK scores (likely to be low in those without CAC). However, there was very poor sensitivity in patients with SLE as well as controls. SLECRISK performed better than the ASCVD Risk Score (correctly risk-stratified 6 more patients; sensitivity of 30.8%) but still yielded limited diagnostic accuracy. We found a low NPV in the ASCVD-stratified SLE group (72.1% vs 89.1% in controls). The NPV improved only slightly with use of SLECRISK (75.7%). Risk analysis (both ASCVD Risk Score and SLECRISK score) of patients with SLE aged > 40 years revealed only modest improvements in sensitivity and poorer NPV, and 12.2% of patients with SLE aged < 40 years of age had abnormal CAC despite being categorized as low ASCVD risk. Overall, these findings suggest that additional screening in this population, which tends to be younger than in traditional risk score cutoffs, may be necessary to avoid incorrectly classifying patients with abnormal CAC as low risk.
Accurate CVD risk assessments and screenings are necessary to inform the appropriate use of statins. Statins have been found to lower LDL levels in the general population23-25 and in people with SLE,26,27 thereby decreasing CVD morbidity and mortality. A retrospective cohort study of SLE patients with hyperlipidemia (n = 4095) found that patients on statins had statistically significantly lower rates of coronary artery disease, endstage renal disease, and all-cause mortality.28 A randomized controlled trial (n = 60) comparing patients with SLE taking atorvastatin vs placebo reported a statistically significant reduction in CAC progression.29
Current guidelines per the US Preventive Services Task Force recommend statin initiation in adults between aged 40 and 75 years who have a 10-year risk of cardiovascular event > 7.5% (C-level recommendation) or > 10% (B-level recommendation).30 However, conventional CVD risk calculators, such as the FRS, ASCVD (AHA/ACC) Risk Score, Systematic Coronary Risk Evaluation (SCORE), and QRISK, frequently underestimate risk in the SLE population.8,31-33 Moreover, the updated ASCVD Risk Score (ACC/AHA 2018) recommends evaluating for risk modifiers, such as autoimmune diseases like SLE and rheumatoid arthritis, to guide statin initiation for primary CVD prevention. Yet, there remains no validated, numerical weighted risk estimates, and these individual risk enhancers have been found to provide only marginal benefit.34 Despite the benefit of statins, a cross-sectional study noted that only 60% of patients with SLE received statins for secondary prevention, and although 43% to 45% of patients were eligible for primary prevention statins, only 87% received them.35 Our data also demonstrate low statin prescription for primary prevention.
The lack of validated numerical estimates has paved the way for SLE-specific risk scores such as QRISK3 or SLECRISK. Whereas these risk modifiers might improve sensitivity, they may not adequately improve accuracy of prediction. A retrospective cohort study (n = 905) found that doubling the FRS increased the sensitivity by more than 2-fold (13.0, 95% CI 0-26.6 to 31.5, 95% CI 13.3-49.7) but still yielded overall poor sensitivity.36 In the development cohort, the SLECRISK score had superior sensitivity for predicting 10-year risk for major adverse cardiovascular events compared to the ACC/AHA model (0.74, 95% CI 0.65-0.83 vs 0.38, 95% CI 0.28-0.48), though this was not externally validated.18 Our study suggests that the SLECRISK score is an improvement, but it still has insufficient sensitivity and NPV to rule out the presence of CAC.
Underestimation of CVD risk in patients with SLE is likely related to the unique risk factors in this population. Among patients with SLE with low ASCVD risk, abnormal CAC was significantly associated with age, lower GFR, longer disease duration, higher HOMA-IR score, and higher LDL and cholesterol levels. This suggests that both traditional and SLE-specific risk factors play a role. Other studies have also found that those with lupus nephritis, kidney disease, or longer disease duration have an increased CVD risk.37-39
The low sensitivity of risk scores, conventional and modified, suggests a need for more accurate CVD screening measures. Electron beam CT (EBCT), a validated imaging modality to quantify CAC, is one possibility.40,41 Although there are limited SLE-specific data describing the association between CAC and adverse cardiac events, CAC has been found to be a good surrogate marker for atherosclerosis and a prognosticator of future cardiac events and mortality in the general population.40,42,43 Using EBCT to determine CAC has been found to effectively modify predicted CVD risk and has been considered as a tool for primary prevention.44 More specifically, the quantification of CAC has been used to effectively identify which patients would benefit from statin initiation for the primary prevention of CVD, mirroring the purpose of cardiovascular risk score calculators.45
At present, the ACC/AHA Cholesterol Guidelines (2018) suggest that CAC screening can be considered in adults (40-75 years old) without diabetes mellitus and with LDL (70-189 mg/dL) and an intermediate ASCVD risk (7.5-20%).46 If CAC is 1-99 and there are no other strong CVD risk factors, statin initiation can be considered; if CAC is > 100, statin therapy is indicated.46 A large prospective cohort study found CAC to be superior to individual risk enhancers (per the 2018 ACC/AHA ASCVD score) in predicting CVD and stratifying intermediate-risk patients.34 The European Society of Cardiac Radiology and the North American Society for Cardiovascular Imaging similarly recommends using CAC to better evaluate those with intermediate CVD risk.47 Major global coronary artery calcium guidelines describe use of CAC screening in asymptomatic patients aged > 40 years with intermediate CVD risk.48 CAC imaging has therefore been considered as an additional modality to differentiate people with intermediate risk.
There are no SLE-specific CAC guidelines. The European Alliance of Associations for Rheumatology (EULAR) currently recommends screening and management of CVD risk factors according to general population guidelines for the SLE patient population and with consideration of SLE-specific risk factors such as disease activity and prolonged corticosteroid use.49 Given the lack of validated and effective risk score calculators, EULAR does not recommend a specific one, but rather recommends comprehensive assessments and preventive management.50 Our study, along with prior literature, suggests that CAC is much more prevalent within the SLE population than may be appreciated using traditional risk prediction scores. This is especially pertinent considering that many SLE patients with CAC had risk prediction scores that were well below the intermediate threshold (ASCVD 7.5-20%) recommended for CAC screening. These data suggest that more aggressive use of CAC scoring may be useful even in low-risk patients in terms of informing statin initiation.
Our study was limited by several factors. First, we conducted a secondary analysis of a previous observational study. Participants were recruited from a single academic health center in the northeastern US, affecting generalizability. Patients who chose to participate in the study could have had a greater interest in CVD due to either existing disease, family history, or exposure, resulting in selection bias. However, only 15% of the SLE group had a previous MI or stroke, a rate comparable to other population-based studies. Moreover, the age of the data may mean that more recent practice patterns are not reflected here, particularly in terms of active statin use. The relatively small sample sizes of the various subgroups could have increased the probability of a type II error for the identification of important risk factors. We additionally made several assumptions to address missing data. However, we had most data available and the relative sensitivity and specificity of both the ASCVD and SLECRISK scores were comparable to those of other studies. We also used CAC as a surrogate for adverse cardiovascular events, which is the main outcome of interest for CVD risk calculators, as real-world CVD event data were unable to be obtained. Although not everyone with CAC experiences a stroke or MI, CAC has been associated with adverse cardiovascular outcomes and is widely believed to represent an indication for statin use.
In conclusion, people with SLE have higher rates of CAC compared to controls, even among those with low predicted risk. Conventional risk calculators such as the ASCVD (ACC/AHA) underestimate the likelihood of CVD, as represented by CAC, with only marginal improvement using modified scores such as the SLECRISK. As such, the use of EBCT to identify CAC, even in patients who appear low risk, might serve as a valuable screening modality to facilitate clinical decision making.
Footnotes
CONTRIBUTIONS
HP: performed research, collected data, analyzed and interpreted data, performed statistical analysis, wrote the manuscript. SS, JVF: performed research, collected data, analyzed and interpreted data, performed statistical analysis. JFB: designed research, performed research, collected data, analyzed and interpreted data, performed statistical analysis, wrote the manuscript.
FUNDING
JFB receives support from 2 Rehabilitation Research & Development Merit Awards (I01 RX003644; I01 RX004770). JFB receives research funding from VA Rehabilitation Research & Development (I01 RX003644). The contents of this work do not represent the views of the Department of Veterans Affairs or the US Government.
COMPETING INTERESTS
JFB has received consulting fees from Formation Bio, CorEvitas, and Burns White. All other authors have no conflicts of interest relevant to this article.
ETHICS AND PATIENT CONSENT
This is a post hoc analysis of a prior study approved by the University of Pennsylvania Institutional Review Board.
- Accepted for publication October 30, 2025.
- Copyright © 2026 by the Journal of Rheumatology
