Cardiovascular events and risk in patients with systemic lupus erythematosus: Systematic literature review and meta-analysis

Abstract

Background

Systemic lupus erythematosus (SLE) is an autoimmune disease that typically affects women aged 16–55 years. Cardiovascular disease (CVD) is a well-recognized complication of SLE. This systematic literature review and meta-analysis evaluated the relative risk (RR; compared with non-SLE controls), absolute risk (AR; as incidence proportion, n/N), and incidence rate (IR) of CVD events (including stroke, myocardial infarction [MI], and CVD [composite or undefined]) in adult patients with SLE. The RR of CV risk factors (including hypertension, diabetes, and metabolic syndrome [MetS]) was also examined.

Methods

PubMed and Embase were searched on September 10, 2020. Observational studies published between January 2010 and September 2020 that reported RR, AR, and/or IR of CVD events, or RR of CV risk factors, were eligible. Pooled risk estimates were calculated using a random-effects model.

Results

Forty-six studies (16 cross-sectional, 15 retrospective cohort, 14 prospective cohort, and 1 case–control) were included in meta-analyses. Most studies were considered high quality (Critical Appraisal Skills Programme checklists). Compared with adults without SLE, patients with SLE had statistically significantly higher RRs (95% CIs) of stroke (2.51 [2.03–3.10]; 12 studies), MI (2.92 [2.45–3.48]; 11 studies), CVD (2.24 [1.94–2.59]; 8 studies), and hypertension (2.70 [1.48–4.92]; 7 studies). RRs of diabetes (1.24 [0.78–1.96]; 3 studies) and MetS (1.49 [0.95–2.33]; 7 studies) were elevated but not significant. RRs of stroke and MI were generally higher in younger versus older patients with SLE. In patients with SLE, the pooled estimate of AR (95% CI) was 0.03 (0.02–0.05), 0.01 (0.00–0.02), and 0.06 (0.03–0.10) for stroke (7 studies), MI (6 studies), and CVD (8 studies), respectively. The pooled estimate of IR per 1000 person-years (95% CI) was 4.72 (3.35–6.32), 2.81 (1.61–4.32), and 11.21 (8.48–14.32) for stroke (10 studies), MI (6 studies), and CVD (8 studies), respectively. Although heterogeneity (based on I² value) was high in most analyses, sensitivity analyses confirmed the robustness of the pooled estimates.

Conclusions

This meta-analysis found an increased risk of stroke, MI, CVD, and hypertension in patients with SLE compared with the general population, despite substantial heterogeneity across the included studies.

Introduction

Systemic lupus erythematosus (SLE) is a systemic autoimmune disease characterized by a wide spectrum of clinical manifestations and an unpredictable relapsing-remitting course. SLE typically affects women aged between 16 and 55 years and is characterized by multiorgan damage mediated by tissue-binding autoantibodies and immune complexes, leading to acute and chronic inflammation of various organs and tissues.^1–3

Cardiovascular disease (CVD) is a well-recognized complication of SLE, and there is strong evidence to suggest that patients with SLE have at least a 2- to 3-fold elevated risk of myocardial infarction (MI), congestive heart failure, cerebrovascular disease, and overall CVD mortality than the general population.^4,5 Patients with SLE have increased mortality, with accumulation of organ damage being the major determinant. A bimodal mortality pattern has been described showing that early mortality is associated with disease activity and infections, whereas late mortality is associated with atherosclerotic complications.⁶

There have been significant improvements in long-term survival in patients with SLE. Morbidity and mortality due to comorbid causes are declining⁷; however, the decline is similar to that of the general population, and mortality remains disproportionately high in the SLE population, particularly in younger patients with SLE.⁸ The relative risk (RR) of CVD is higher in younger women (aged 19–39 years) with SLE compared with the general population,⁹ and the absolute risk (AR) of CVD increases with age in both men and women with SLE.¹

Understanding the magnitude of CVD risk, including stroke, MI, and cardiovascular (CV) death, as well as traditional CV risk factors, in patients with SLE would support the identification of high-risk patients in clinical practice and allow for implementation of tailored prevention and treatment strategies in addition to SLE management. The objective of this meta-analysis was to evaluate the RR (compared with the general population or suitable proxies), AR, and incidence rate (IR) of CV events, including stroke, MI, and CVD, in adult patients with SLE. An additional objective was to evaluate the ratio of risk of CV risk factors, including hypertension, diabetes, and metabolic syndrome (MetS), in patients with SLE compared with the general population.

Methods

Search strategy

This study was conducted using the “Preferred Reporting Items for Systematic Reviews and Meta-analysis” guidelines¹⁰ and followed a prespecified protocol (not registered). For the systematic literature review (SLR) and meta-analysis (MA), the available literature was retrieved from the electronic medical databases MEDLINE (using the PubMed platform) and Embase (using the Elsevier platform) on September 10, 2020. The search was limited to observational studies published between January 2010 and September 2020 that reported on CV events and CV risk factors in adult patients with SLE. The keywords used in the search focused on SLE, CV events, CV risk factors, and the risk or incidence of CV events (Supplementary Table 1, Supplementary Table 2). The CV events of interest included MI, angina, heart failure, stroke, peripheral arterial disease, transient ischemic attack (TIA), major adverse CV events, and CVD. The CV risk factors included hypertension, obesity, smoking, diabetes, dyslipidemia, MetS, and chronic kidney disease.

Selection criteria

Although there were no geographic or language limitations on the searches, only English-language articles were included during screening. Observational studies (cohort, cross-sectional, case–control, registry studies, and other observational) were eligible for inclusion; randomized and nonrandomized clinical trials, case reports, editorials, etc., were excluded. SLRs and MAs were hand-searched to identify additional relevant studies. Studies that included patients with diseases other than SLE or pediatric (aged <18 years) patients with SLE were not eligible, unless results for adults with SLE were reported separately. The outcomes of interest included the RR of CV events in adults with SLE compared with adults without SLE; the AR of CV events in patients with SLE; and the IR of CV events in patients with SLE. In addition, to evaluate the ratio of risk of CV risk factors, studies reporting on CV risk factors in patients with SLE compared with the general population were also included in the SLR. Subgroups of interest included patients differentiated by age, sex, ethnicity, geographical location, disease severity, disease duration, prior treatment received, and antiphospholipid syndrome/antiphospholipid antibodies. The CV events reported in this publication were stroke, MI, and general CVD (e.g., composite outcomes or CVD not otherwise defined), and the CV risk factors were hypertension, diabetes, and MetS. Other events included in the SLR were not prioritized for the MA because of insufficient data.

Data extraction and quality assessment

Data were extracted from full-text publications. References to other publications within a study were traced to original sources, where appropriate. When multiple publications reporting on the same study were identified, the primary publication was extracted first; data from secondary publications were extracted if not reported in the primary publication. All extracted data were quality-checked and verified against the source document by a second researcher not involved with the extraction. A quality assessment recommended by the Critical Appraisal Skills Programme (CASP; https://casp-uk.net/) was conducted for each individual observational study. Individual CASP checklists were used to evaluate case–control studies and cohort studies. A modified version of the CASP cohort study checklist was used to evaluate cross-sectional studies. Studies were classified as low, medium, or high quality, or as unclear; all studies were included in primary analyses regardless of quality.

Statistical analysis

General meta-analysis methods

Meta-analyses for RR, AR (incidence proportion), and IR were conducted for each outcome when there were ≥3 studies reporting the respective measure for the respective outcome. The restricted maximum likelihood random-effects model was used to calculate the pooled risk estimates and 95% confidence intervals (CIs) for all outcomes. Fixed-effects models were also fit for completeness. Higgins’s I² was used to estimate the percentage of variance explained by heterogeneity.¹¹ Sensitivity analyses were performed to assess the robustness of the primary pooled estimates as well as the impact of study design or risk measures on the effect size; these included leave-one-out analysis (i.e., MA in which exactly one study is omitted; this method demonstrates how each individual study affects the overall estimate), limiting analysis to studies with comparable CV event definitions, and limiting analysis to high-quality studies. Metaprop or metafor in R (R Foundation for Statistical Computing, Vienna, Austria) was used for analyses.¹² Forest plots were visually inspected to assess the consistency of the study data.

Meta-analysis methods for relative effects

Measures such as odds ratio, hazard ratio (HR), RR, IR ratio, or standardized IR were used for comparative measures of the ratio of risk of stroke, MI, CVD, hypertension, diabetes, and MetS for patients with SLE versus the general population. Because data were scarce for most endpoints, all comparable studies were included in the primary MA. These measures were pooled in the MA and are hereafter referred to collectively as RRs. Because of differences in adjustment factors used across the studies, unadjusted measures were used in preference to adjusted measures where both were available. Studies were included in the analysis regardless of the length of follow-up under the assumption of proportional hazards over time. Subgroup analyses were based on age and sex compared with matched non-SLE controls.

Meta-analysis methods for absolute risk

For the analysis of AR of stroke, MI, and CVD in patients with SLE, the AR was defined as a proportion. The contributing studies provided information on the number of events (n/N), and the double-arcsine transformation was used for this analysis.

Meta-analysis methods for incidence rates

For the analysis of IR of stroke, MI, and CVD in patients with SLE, results from studies were adjusted to be in the consistent form of per 1000 person-years (PY). The metafor package was used with the metarate command, with the events based on the IR per 1000 PY and the PY of follow-up used as inputs.

Results

Search results

The electronic literature search identified 1591 articles for title and abstract screening, of which 278 articles were retrieved for Level 2 screening of the full text; 7 additional articles were identified from published SLRs (Figure 1). A total of 87 articles met the inclusion criteria for the SLR. After review, 46 of these were included in ≥1 MA (Figure 1), and 41 were excluded for various reasons (Supplementary Table 3).

Figure 1.

Flow diagram of the study selection process. MA: meta-analysis; SLR: systematic literature review.

Characteristics of included studies

The 46 studies in the MA included 16 cross-sectional studies, 15 retrospective cohort studies, 14 prospective cohort studies, and 1 case–control study (Table 1). Based on CASP checklists, 36 studies were deemed to be of high quality (although none met all criteria uniformly), 8 studies were medium quality, and 1 study was low quality; the quality of 1 study was unclear. Studies were conducted in a broad range of countries from Europe, North America, South America, Asia, and the Middle East. In most studies, patients were identified as having SLE based on American College of Rheumatology criteria or International Classification of Diseases codes.

Table 1.

Baseline characteristics of the included studies and participants.

Study (county)	Year	Patient population	SLE criteria	Reference population	Study design		Length of follow-up and/or dates	Quality appraisal a	Endpoint measured
Aljohani et al.13 (Canada)	2017	SLE patients (n = 255)	≥4 ACR criteria or 3 ACR criteria plus histologic lesion of SLE on renal or skin biopsy	NA	Prospective cohort		5 years (starting between 1969 and 2010)	High	MI and CVD (angina, CABG, MI, cardiomyopathy, valvular disease, and pericarditis)
Arkema et al.14 (Sweden)	2017	Adult patients with incident SLE (n = 3390)	≥2 discharge diagnoses listing an SLE ICD code, and ≥1 registered with a relevant specialist	General population matched on age, sex, and county (n = 16,730)	Prospective cohort		2003–2013	High	Stroke
Aviña-Zubieta et al.15 (Canada)	2017	Adult patients with incident SLE (n = 4863)	ICD-9-CM SLE code by a rheumatologist or from a hospital (710.0 or ICD-10 M32.1, M32.8, and M32.9) or 2 SLE ICD-9-CM codes (710.0) ≥2 months apart within 2 years by a nonrheumatologist physician	General population matched on age, sex, and time (n = 49,316)	Prospective cohort		1996–2010	High	Stroke, MI, and CVD (stroke and MI)
Barbhaiya et al.16  (USA)	2017	Adults with SLE (n = 65,788)	≥3 SLE ICD-9 codes (710.0), from hospital discharge diagnoses or physician visit claims, ≥30 days apart	NA	Prospective cohort		2000–2010	High	CVD (stroke and MI)
Barbhaiya et al.17 (USA)	2019	Adults with SLE (n = 65,788)	≥3 SLE ICD-9 codes (710.0), from hospital discharge diagnoses or physician visit claims, ≥30 days apart	NA	Prospective cohort		2000–2010	High	Stroke
Barbhaiya et al.18  (USA)	2020	Adult patients with SLE (n = 40,212)	≥3 SLE ICD-9 codes (710.0), from hospital discharge diagnoses or physician visit claims, ≥30 days apart	General population matched for age and sex (n = 160,848)	Prospective cohort		2007–2010	High	Stroke, MI, and CVD (stroke and MI)
Barnado et al.19 (USA)	2018	Patients with SLE (n = 1097)	≥4 SLE ICD-9 codes (710.0) and a positive ANA with titer ≥1:160 while excluding ICD-9 codes for systemic sclerosis (710.1) and dermatomyositis (710.3)	Age, sex, and race matched to SLE patients (n = 1425)	Retrospective cohort		Black or African American: 9 years White: 10 years (dates NR)	High	Hypertension
Bartels et al.20 (USA)	2014	Incident cases of SLE (n = 70)	ICD-9 SLE diagnosis codes (710.0)	Population matched by age and sex (n = 2565)	Retrospective cohort		1991–2009	High	Stroke, MI, and CVD (MI, stroke, CHF, and ischemic heart disease)
Bengtsson et al.21 (Sweden)	2012	Prevalent SLE patients (n = 277)	≥4 of the 1982 ACR classification criteria and the updated criteria	Age- and sex-matched controls (n = 516)	Prospective cohort		7 years (2000–2007)	High	Stroke, MI, CVD (stroke and MI), and hypertension
Bhat et al.22 (India)	2015	Patients with SLE (n = 82)	1997 ACR classification criteria	Age- and sex-matched healthy controls (n = 82)	Cross-sectional		2012–2013	Medium	Hypertension and MetS
Chiu et al.23  (Taiwan)	2012	Newly diagnosed SLE patients (n = 11,637)	Catastrophic illness registration cards for SLE (ICD-9-CM codes 710.0). SLE diagnosis established by rheumatologists based on ACR 1997 revised criteria	Population without SLE and matched for age, sex, and comorbidities (n = 58,185)	Prospective cohort		2000–2007	High	Stroke
Chou et al.24 (Taiwan)	2014	Diagnosed SLE patients (n = 10,954)	NR	General population (n = 43,816)	Retrospective cohort		1998–2008	High	MI
Dhar et al.25 (USA)	2013	Premenopausal women with SLE (n = 677)	NR	General population matched for age group and sex (n = NR)	Retrospective cohort		NR	High	MI, hypertension, diabetes
Demir et al.26 (Turkey)	2016	SLE patients without MetS (n = 250)	≥4 ACR classification criteria	NA	Cross-sectional		NR	Medium	CVD (CAD, cerebrovascular events including MI and stroke)
Dregan et al.27 (UK)	2017	Adults with SLE (n = 654)	NR	Adults without a diagnosis of a chronic inflammatory disorder (n = 483,559)	Cross-sectional		NR	High	Stroke, diabetes
Falasinnu et al.28 (USA)	2019	Adult patients with SLE (n = 1036)	≥2 clinic visits with an ICD-9 code of 710.0 or an ICD-10 code of M32.1, M32.8, and M32.9	Non-SLE controls matched for age, sex, and service date (n = 8875)	Prospective cohort		2005–2016	High	CVD (ICD-9 390–459 and ICD-10 I00-I99 [diseases of the circulatory system])
Fatemi et al.29 (Iran)	2018	Patients with SLE and MetS (n = 35)	≥4 criteria from revised ACR criteria	Controls attending hospital for a regular check-up, with MetS (n = 28)	Cross-sectional		NR	High	MetS
Fernández-Garcés et al.30 (Spain)	2019	Women with SLE (n = 335)	≥4 criteria from revised ACR criteria	NA	Cross-sectional		2003–2013	High	Stroke and CVD (ischemic heart disease [including MI and angina], stroke, and PAD)
Fernández-Gutiérrez et al.31 (Spain)	2017	Patients with SLE (n = 760)	1997 ACR criteria	Patients with other inflammatory diseases (n = NR)	Cross-sectional		2007–2010	High	CVD (angina, MI, peripheral vascular disease, and stroke)
Fernández-Nebro et al.32 (Spain)	2015	SLE patients (n = 3658)	1997 ACR criteria	NA	Cross-sectional		2011–2012	High	Stroke and MI
García-Villegas et al.33 (Mexico)	2015	Premenopausal women with SLE (n = 238)	≥4 ACR criteria	NA	Prospective cohort		2001–2008	High	Stroke, MI, and CVD (MI, stroke, arterial or venous thrombosis, and DVT)
Guraieb-Chahín et al.34 (Mexico)	2020	Patients newly diagnosed with SLE (n = 11,288)	ACR criteria	NA	Retrospective cohort		2001–2012	Medium	Stroke
Hassan et al.35 (Saudi Arabia)	2013	Patients with SLE (n = 120)	ACR criteria	Controls with no inflammatory arthritis or cardiovascular diseases (n = 100)	Case–control		NR	High	Diabetes
Hatef-Fard et al.36 (Iran)	2016	Patients with SLE (n = 150)	ACR criteria	Healthy controls (n = 220)	Cross-sectional		NR	Low	MetS
Hermansen et al.37 (Denmark)	2017	Patients diagnosed with SLE without LN (n = 1411)	ICD-10 codes DM321, DM328, and DM329	Matched population controls (n = 7055)	Retrospective cohort		Median (IQR) (years): 7.8 (3.3–12.7) (1995–2011)	High	Stroke and MI
Hesselvig et al.38 (Denmark)	2017	Patients with SLE (n = 3747)	ICD-10 code M32	General population controls (n = 5,513,739)	Prospective cohort		1997–2011	High	CVD (stroke, MI, and CV death)
Kim et al.39 (USA)	2017	Patients with SLE (n = 95,400)	SNOMED-CT term for SLE	Nondeceased inpatients and outpatients included in hospital databases (n = 45,189,140)	Retrospective cohort	1999–2016		High	MI and hypertension
Kuo et al.40 (UK)	2019	Incident SLE cases (n = 1605)	Read codes	Matched controls (n = 6284)	Retrospective cohort	1997–2005		High	MI
Li et al.41 (USA)	2020	Adults with incident SLE (n = 15,120)	≥3 ICD-9 SLE codes (710.0), from hospital discharge diagnoses or physician visit claims, ≥30 days apart	NA	Prospective cohort	2000–2010		High	CVD (MI, stroke, PCI, heart failure, and CABG)
Lim et al.42 (Korea)	2018	SLE patients (n = 18,575)	ICD-10 code M32	Individuals with no SLE matched by age and sex (n = 92,875)	Retrospective cohort	2008–2014		High	Stroke and MI
Lin et al.43 (Taiwan)	2014	Newly diagnosed adult SLE patients (n = 1207)	ICD-9-CM code 710.0	Matched population (n = 9656)	Retrospective cohort	2000–2008		High	MI
Liou et al.44 (Taiwan)	2014	SLE patients (n = 621)	ICD-9-CM code 710.0 with diagnosis by rheumatologist using ACR criteria	Age-matched controls (n = 2484)	Retrospective cohort	2004–2007		High	Stroke
Magder and Petri45 (USA)	2012	Patients with SLE (n = 1874)	NR	NA	Prospective cohort	1987–2010		High	CVD (stroke, MI, angina, PCI, CABG, and claudication)
Medeiros et al.46 (Brazil)	2016	Patients with SLE (n = 146)	1997 ACR criteria	Sex-, age-, ethnicity-, and education-matched controls (n = 101)	Cross-sectional	NR		Medium	MetS
Mobini et al.47 (Iran)	2018	Patients with SLE (n = 73)	2012 SLICC criteria	Healthy controls without rheumatologic disorders (n = 73)	Cross-sectional	NR		Medium	MetS
Mok et al.48 (China)	2018	SLE patients (n = 577)	≥4 ACR criteria	NA	Retrospective cohort	Mean: 66.3 months (starting 2010)		High	CVD (MI, ischemic cardiomyopathy, and aortic dissection)
Montes et al.49 (Brazil)	2016	SLE patients (n = 105)	SLICC criteria	NA	Cross-sectional	Mean: 8.98 years (starting 2013–2014)		Medium	MI
Muniz et al.50 (Brazil)	2015	Patients with SLE (n = 103)	≥4 ACR criteria	Healthy controls without autoimmune diseases (n = 35)	Cross-sectional	2009–2011		High	MetS
Parker et al.51 (UK)	2011	Women with SLE (n = 200)	≥4 1997 ACR criteria or 3 criteria in the absence of an alternative diagnosis	Healthy females free from any chronic inflammatory disease matched on ethnicity and same geographical area (n = 100)	Cross-sectional	NR		High	Hypertension and MetS
Rees et al.52 (UK)	2016	Prevalent SLE cases (n = 7732)	Read codes	Matched controls based on age, sex, and practice (n = 28,079)	Retrospective cohort	1999–2012		High	Stroke and CVD (definition NR)
Sabio et al.53 (Spain)	2011	Women with SLE (n = 112)	≥4 ACR criteria	Healthy age-matched women (n = 223)	Cross-sectional	NR		Unclear	Hypertension
Shaharir et al.54 (Malaysia)	2019	SLE patients (n = 418)	≥4 1997 ACR criteria, plus SLICC criteria for patients after 2012	NA	Cross-sectional	2016–2017		Medium	Stroke and CVD (stroke and ischemic heart disease)
Tselios et al.55 (Canada)	2017	Women with SLE (n = 210)	1997 ACR criteria or 3 criteria and a confirmatory skin or renal biopsy	Age-matched healthy women (n = 138)	Prospective cohort	15 years (2000–2015)		High	CVD (angina, MI, TIA, stroke, and CV death)
Wang et al.56 (Taiwan)	2012	Newly diagnosed SLE (n = 13,689)	ICD-9-CM code 710.0	No SLE controls matched by age, sex, and index year (n = 54,756)	Retrospective cohort	Mean: 5.47 years (1997–2008)		High	Stroke
Wang et al.57 (China)	2012	Patients with SLE (n = 1072)	≥4 ACR criteria	NA	Cross-sectional	2006–2008		Medium	CVD (stroke, MI, angina, peripheral vascular damage, CABG, PCI, TIA, and CHF)
Yang et al.58 (Taiwan)	2019	Newly diagnosed with SLE receiving HCQ medication (n = 826)	ICD-9-CM code 710.0	NA	Retrospective cohort	2000–2013		High	CVD (stroke and CAD)

ACR: American College of Rheumatology; CABG: coronary artery bypass grafting; CAD: coronary artery disease; CHF: congestive heart failure; CV: cardiovascular; CVD: CV disease; DVT: deep vein thrombosis; HCQ: hydroxychloroquine; ICD: International Classification of Diseases; ICD-9-CM: ICD, Ninth Revision, Clinical Modification; ICD-10: ICD, Tenth Revision; IQR: interquartile range; LN: lupus nephritis; MetS: metabolic syndrome; MI: myocardial infarction; NA: not applicable; NR: not reported; PAD: peripheral artery disease; PCI: percutaneous coronary intervention; SLE: systemic lupus erythematosus; SLICC: Systemic Lupus International Collaborating Clinics; SNOMED-CT: Systemized Nomenclature of Medicine-Clinical Terms; TIA: transient ischemic attack.

^aStudy quality was assessed using Critical Appraisal Skills Programme (CASP; https://casp-uk.net/) checklists for observational studies.

The CV events of stroke, MI, and CVD were reported as endpoints in 18, 15, and 20 studies, respectively (Table 1). The definition of CVD varied across studies, although most studies included key events such as MI and stroke within a composite CVD outcome. The CV risk factors of hypertension, diabetes, and MetS were reported as endpoints in 7, 3, and 7 studies, respectively. Given the limited number of studies reporting other CV events and CV risk factors (e.g., smoking and dyslipidemia), the MA focused on the CV events of stroke, MI, and CVD, together with the CV risk factors of hypertension, diabetes, and MetS.

Relative risk of stroke, MI, and CVD in patients with SLE compared with the general population

Twelve studies were included in the analysis of the RR of stroke in patients with SLE compared with the general population.^{14,15,18,20,21,23,27,37,42,44,52} A statistically significantly higher risk of stroke was observed in those with SLE than in the general population (Figure 2(a)). The RR estimates across the studies ranged from 1.61 to 5.34, and the pooled estimate was 2.51 (95% CI: 2.03–3.10). Substantial heterogeneity was observed in the RR estimates across the studies (I² = 89.6%). As there were differences in the measures reported by most studies, a sensitivity analysis was conducted including only those studies reporting an HR (8 studies).^{14,15,18,23,37,42,44,56} The pooled estimate was similar to the primary analysis (2.39; 95% CI: 1.96–2.92), although substantial heterogeneity remained (I² = 90.5%; Supplementary Figure 1). Leave-one-out sensitivity analyses of studies reporting HR resulted in estimates for the RR of stroke ranging from 2.37 to 2.83 (Supplementary Table 4).

Figure 2.

RR of (a) stroke, (b) MI, and (c) CVD in patients with SLE versus the general population. CI: confidence interval; CVD: cardiovascular disease; df: degrees of freedom; FE: fixed-effects; MI: myocardial infarction; RE: random-effects; RR: relative risk; SLE: systemic lupus erythematosus.

In total, 11 studies were included in the primary analysis of the RR of MI in patients with SLE compared with the general population (Figure 2(b)).^{15,18,20,21,24,25,39,40,42,43} A statistically significant, 2.9-fold higher risk of MI was seen in those with SLE compared with the general population (RR = 2.92 [95% CI: 2.45–3.48]), with RRs from individual studies ranging from 1.71 to 5.50. Substantial heterogeneity was observed (I² = 86.7%), possibly related to differences in the measures of RR used in the included studies. To investigate this, a sensitivity analysis including only those studies reporting an HR (n = 7) was conducted,^{15,18,24,37,40,42,43} resulting in a pooled estimate of 2.76 (95% CI: 2.46–3.10), similar to the primary analysis, but with reduced heterogeneity (I² = 48.3%) (Supplementary Figure 2).

Eight studies were included in the primary analysis of the RR of CVD in patients with SLE compared with the general population (Figure 2(c)).^{15,18,20,21,28,38,52,55} A statistically significant, 2.2-fold higher risk of CVD was observed in those with SLE compared with the general population (RR = 2.24 [95% CI: 1.94–2.59]). Substantial heterogeneity (I² = 73.0%) was observed in the RR estimates across the studies, which ranged from 1.27 to 3.70. A sensitivity analysis using only those studies with CVD definitions of stroke and/or MI (n = 4)^15,18,21,38 found a pooled estimate of 2.10 (95% CI: 1.63–2.70), although substantial heterogeneity remained (I² = 83.7%) (Supplementary Figure 3). To evaluate whether differing measures of risk contributed to heterogeneity, a sensitivity analysis using only those studies that reported HR (n = 4)^15,18,38,55 found a pooled estimate of 2.34 (95% CI: 2.00–2.73), but substantial heterogeneity remained (I² = 71.6%) (Supplementary Figure 4).

Relative risk of hypertension, diabetes, and MetS in patients with SLE compared with the general population

The RR of hypertension was statistically significantly higher in patients with SLE compared with the general population (2.70 [95% CI: 1.48–4.92]; 7 studies; Supplementary Figure 5(a)).^{19,21,22,25,39,51,53} Heterogeneity was high (I² = 99.7%); leave-one-out sensitivity analyses confirmed the estimated RR (range: 2.00–3.13), but substantial heterogeneity remained (I² = 90.6–99.8%) (Supplementary Table 5). Although not statistically significant, the RRs of diabetes (1.24 [95% CI: 0.78–1.96]; 3 studies; Supplementary Figure 5(b))^25,27,35 and MetS (1.49 [95% CI: 0.95–2.33]; 7 studies; Supplementary Figure 5(c))^{22,29,36,46,47,50,51} also indicated an elevated risk of these factors in patients with SLE compared with non-SLE controls. Heterogeneity was moderate in the diabetes analysis (I² = 30.4%) and substantial in the MetS analysis (I² = 69.6%); leave-one-out sensitivity analysis that excluded 1 study³⁶ with a notably lower RR resulted in an estimated MetS RR of 1.74 (95% CI: 1.32–2.30), with I² = 0%.

AR of stroke, MI, and CVD in patients with SLE

In total, 7 studies were included in the primary analysis of the AR (i.e., incidence proportion) of stroke in patients with SLE (Figure 3(a)).^{18,30,32,33,42,44,54} The AR of stroke in patients with SLE ranged from 0.01 to 0.06, and the pooled estimate was 0.03 (95% CI: 0.02–0.05). Substantial heterogeneity was observed in the AR estimates across the studies (I² = 98%). The results of the leave-one-out sensitivity analysis resulted in estimates ranging from 0.03 to 0.04 (Supplementary Table 6). Because the heterogeneity could be due to differences in the definition of stroke, a sensitivity analysis was conducted using only those studies where stroke definitions were reported and were not likely to include TIAs (n = 3).^18,30,33 This estimate resulted in a pooled AR of 0.03 (95% CI: 0.01–0.06), although substantial heterogeneity remained (I² = 94%) (Supplementary Figure 6). A sensitivity analysis was also conducted to evaluate the impact of disease duration on the AR of stroke, which included 4 studies reporting disease duration.^30,32,44,54 The pooled estimate for this sensitivity analysis was 0.05 (95% CI: 0.04–0.06), with moderate heterogeneity (I² = 57%) (Supplementary Figure 7). Visual assessment suggests a possible trend for higher risk with longer disease duration, but the small number of studies limits interpretation.

Figure 3.

AR of (a) stroke, (b) MI, and (c) CVD in patients with SLE. AR: absolute risk; CI: confidence interval; CVD: cardiovascular disease; FE: fixed-effects; FU: follow-up; MI: myocardial infarction; RE: random-effects; SLE: systemic lupus erythematosus.

In total, 6 studies were included in the primary analysis for the AR of MI in patients with SLE (Figure 3(b)).^{13,18,32,33,42,49} The AR of MI ranged from 0.00 to 0.02, with a pooled estimate of 0.01 (95% CI: 0.00–0.02). Substantial heterogeneity was observed in the AR estimates across the studies (I² = 94%). A leave-one-out sensitivity analysis resulted in estimates of 0.01 for all analyses (Supplementary Table 7).

In total, 8 studies were included in the primary analysis of the AR of CVD in patients with SLE (Figure 3(c)).^{13,18,26,30,33,48,54,57} The AR of CVD ranged from 0.02 to 0.16, with a pooled estimate of 0.06 (95% CI: 0.03–0.10). Substantial heterogeneity was observed in the AR estimates across the studies (I² = 98%). To evaluate whether the heterogeneity could be due to differences in how CVD was defined, a sensitivity analysis was conducted with only those studies whose CVD definitions were stroke and/or MI (n = 4).^18,26,33,54 The pooled estimate for this sensitivity analysis was 0.07 (95% CI: 0.02–0.14), although substantial heterogeneity remained (I² = 98%) (Supplementary Figure 8). To further evaluate the heterogeneity and robustness of the pooled estimates, a sensitivity analysis was also conducted with only those studies assessed as high quality (n = 5).^{13,18,30,33,48} The pooled AR estimate for this sensitivity analysis was 0.05 (95% CI: 0.02–0.09), with substantial heterogeneity remaining (I² = 95%) (Supplementary Figure 9).

Incidence rate of stroke, MI, and CVD in patients with SLE

Ten studies were included in the primary analysis of IR of stroke in patients with SLE (Figure 4(a)).^{14,15,17,18,34,37,42,44,52,56} The IR of stroke ranged from 1.25 to 8.18 per 1000 PY, with a pooled estimate of 4.72 (95% CI: 3.35–6.32). Substantial heterogeneity was observed (I² = 99%). Pooled estimates from the leave-one-out sensitivity analysis ranged from 4.43 to 5.22 (Supplementary Table 8). Further sensitivity analysis was conducted using only those studies where stroke definitions were reported and were not likely to include TIAs (n = 6)^{14,15,17,18,34,37}; in this analysis, the pooled estimate for the IR of stroke was 4.94 (95% CI: 3.05–7.27), but substantial heterogeneity remained (I² = 99%) (Supplementary Figure 10).

Figure 4.

IR of (a) stroke, (b) MI, and (c) CVD in patients with SLE. CI: confidence interval; CVD: cardiovascular disease; FE: fixed-effects; IR: incidence rate; MI: myocardial infarction; PY: person-years; RE: random-effects; SLE: systemic lupus erythematosus.

Six studies were included in the primary analysis of the IR of MI in patients with SLE (Figure 4(b)).^{15,18,24,37,42,43} The IR of MI ranged from 1.27 to 6.40 per 1000 PY, with a pooled estimate of 2.81 (95% CI: 1.61–4.32). Substantial heterogeneity was observed (I² = 97%). The leave-one-out sensitivity analysis resulted in estimates ranging from 2.25 to 3.20 and with substantial heterogeneity remaining (I² ranging from 95.1% to 97.7%) (Supplementary Table 9).

In total, 8 studies were included in the primary analysis of the IR of CVD in patients with SLE (Figure 4(c)).^{15,16,18,33,41,45,52,58} The IR of CVD ranged from 5.18 to 16.10 per 1000 PY, with a pooled estimate of 11.21 (95% CI: 8.48–14.32). Substantial heterogeneity was observed (I² = 98%). A sensitivity analysis including only those studies with CVD definitions of stroke and/or MI found an estimate of 9.26 (95% CI: 8.92–9.61) with no heterogeneity (I² = 0%) (Supplementary Figure 11). The leave-one-out sensitivity analysis resulted in estimates ranging from 10.51 to 12.21 (Supplementary Table 10).

Subgroup analyses of RR of stroke, MI, and CVD by age and sex

Meta-analyses were not conducted for CV events by age and sex because of limited data. As an alternative, results from individual studies are shown. Although studies used different age group definitions, the RRs of stroke (5 studies; Figure 5(a)),^{14,15,18,37,56} MI (5 studies; Figure 5(b)),^{15,18,21,24,37} and CVD (3 studies; Figure 5(c))^15,18,38 were generally higher in younger patients with SLE compared with older patients with SLE, especially in patients younger than 40 years, when compared with age-matched non-SLE controls.

Figure 5.

RR of (a) stroke by age, (b) MI by age, (c) CVD by age, (d) stroke by sex, and (e) MI by sex. CI: confidence interval; CVD: cardiovascular disease; MI: myocardial infarction; RR: relative risk.

For sex, the RR of stroke in patients with SLE compared with the general population appeared higher in women than in men in 3 of 4 included studies (Figure 5(d)).^14,37,52,56 In contrast, there was no consistent pattern in the RR of MI by sex among 4 included studies (Figure 5(e)).^21,24,37,43 Only 2 studies reported CVD by sex;^31,58 therefore, these studies are not displayed.

Discussion

The main objective of this MA was to provide pooled estimates of the RR of CV events in patients with SLE compared with the general population or suitable proxies. Despite substantial heterogeneity, there was a 2- to 3-fold increased risk of stroke, MI, and CVD with SLE compared with the general population, consistent with previous MAs.^4,5,59,60 Yazdany et al. reported that the RRs of stroke and MI were significantly higher in those with SLE compared with the general population (2.13 [95% CI: 1.73–2.61] and 2.99 [95% CI: 2.34–3.82], respectively).⁵ Similarly, Gu et al. reported an RR of 1.96 (95% CI: 1.52–2.53) for stroke and an RR of 3.04 (95% CI: 1.81–5.11) for MI.⁵⁹ Lu et al. reported an RR of 2.30 (95% CI: 1.52–3.50) for stroke, 2.66 (95% CI: 1.97–3.59) for MI, and 2.35 (95% CI: 1.95–2.84) for CVD,⁴ whereas Restivo et al., who only included prospective cohort studies published between 2015 and 2019, reported that patients with SLE had an elevated risk of symptomatic CV events (RR = 1.98 [95% CI: 1.18–3.31]).⁶⁰ Interestingly, these previous MAs found a similarly high level of heterogeneity as in the current MA. Despite this heterogeneity, the consistency of RR estimates from the primary and sensitivity analyses, and similar findings across the published studies, supports the elevated risk of CV events in patients with SLE. Moreover, visual assessment of the forest plots of subgroup analyses by age suggested that the RR of stroke, MI, and CVD compared with an age-matched non-SLE population was greater in younger patients with SLE compared with older patients with SLE, consistent with the findings of Yazdany et al. for stroke.⁵ Data on the RR of stroke, MI, and CVD in patients with SLE versus sex-matched non-SLE cohorts were limited.

Unlike previous MAs, the current analysis also estimated the AR and IR of CV events in patients with SLE. The estimated pooled ARs for stroke, MI, and CVD were 0.03, 0.01, and 0.06, respectively, which are generally higher than observed in the general population without SLE.^18,42,44 However, differences were seen in the AR across the studies, especially for CVD, which may relate to the diversity of outcomes included in this category. There was a visual trend for increased AR of CVD with longer follow-up, although visual assessment of the data did not indicate this relationship for stroke or MI (data not shown). The pooled estimates for the IR of stroke and MI in this MA (4.72 per 1000 PY and 2.81 per 1000 PY, respectively) were similar to those reported by Gu et al. (4.4 per 1000 PY and 2.0 per 1000 PY, respectively).⁵⁹ As with AR, the IR of these CV events, as well as for CVD (11.21 per 1000 PY in this MA), was higher in patients with SLE than in the general population.^{14,15,18,24,37,42,43,44,52,56}

The RR of the CV risk factors of hypertension, diabetes, and MetS was also elevated in patients with SLE compared with non-SLE controls (2.70 [1.48–4.92], 1.24 [0.78–1.96], and 1.49 [0.95–2.33], respectively), although statistical significance was only observed for hypertension. Our MA results are similar to those of Lu et al., who reported a significantly increased RR for hypertension but not diabetes (2.31 [1.62–3.29] and 1.19 [0.90–1.58], respectively).⁴ Although traditional CV risk factors are likely to contribute to increased risk of CV events in patients with SLE, the underlying mechanisms are complex and are believed to also include inflammation caused by SLE, the presence of antiphospholipid antibodies, and SLE-specific risk factors (e.g., SLE disease activity and duration, renal disease, organ damage [as measured by the Systemic Lupus International Collaborating Clinics/American College of Rheumatology Damage Index], and corticosteroid therapy).^1,61,62

To our knowledge, this is the first SLR and MA to assess the risk of not only stroke and MI but also general CVD, including subgroups based on age and sex, as well as the risk of CVD risk factors such as hypertension, diabetes, and MetS across multiple observational study types in patients with SLE compared with the general population. As noted above, this is also the first MA to report pooled estimates of AR and IR for CV events in patients with SLE. Moreover, this MA applied a single protocol to the investigation of multiple endpoints and epidemiologic measures. An additional strength of this MA is that it uses evidence from multiple observational study types to investigate the risk of CV events in adults with SLE compared with the general population or other control groups. By using multiple study types, the evidence base is widened and more comprehensive. Although most included studies were considered high quality, the potential for bias is a limitation of all observational studies. Moreover, the inclusion of multiple study types likely contributed to the heterogeneity observed for most endpoints. This heterogeneity may have resulted not only from differences in study design but also from differences in patient characteristics, study quality, length of follow-up, and control group selection and matching across the studies, as well as the use of different definitions of events, especially for CVD and, to a lesser extent, stroke. However, although the results from sensitivity analyses of CVD and stroke using more comparable definitions were similar to the primary analysis results, heterogeneity remained, and caution should be used in the interpretation of the findings. Finally, this MA was unable to examine some factors that potentially affect CVD risk in patients with SLE, such as smoking and dyslipidemia, because of a limited number of studies on these factors.

Conclusions

In summary, evidence from this MA suggests that there is an increased risk and incidence of stroke, MI, and CVD in patients with SLE compared with the general population. Despite study differences and heterogeneity in the pooled estimates, the results of sensitivity analyses were similar to those of the primary analyses, supporting the robustness of the estimates of risk. Given the high risk of CV mortality and morbidity in patients with SLE, especially in younger patients, it is imperative that risk factors are routinely assessed and appropriate preventive measures are integrated into the management plan. Additionally, more rigorous and comparable prospective cohort studies should be conducted to further evaluate CV risk and the impact of disease- and patient-related risk factors.

ORCID iD

Natalia Bello https://orcid.org/0000-0001-5042-9460