Anifrolumab Plus Standard of Care Reduces Long‐Term Cardiovascular Damage in Adults With Systemic Lupus Erythematosus: Findings From a Real‐World External Control Study

Abstract

Objective

Despite improvements in overall survival, cardiovascular (CV) events remain a leading cause of death in patients with systemic lupus erythematosus (SLE). Anifrolumab controls disease activity while reducing glucocorticoid use and has been associated with reduced organ‐damage accrual; however, its long‐term efficacy for reducing CV damage is unknown. This study evaluates the effect of anifrolumab plus standard of care (SoC) on long‐term CV damage in SLE compared with real‐world external controls.

Methods

We included 354 patients who initiated anifrolumab 300 mg in the Treatment of Uncontrolled Lupus via the Interferon Pathway (TULIP) trials in the cohort of patients who received anifrolumab. The real‐world SoC cohort comprised 561 patients from the University of Toronto Lupus Clinic registry, selected using matched eligibility criteria. Baseline characteristics were balanced using propensity score weighting. Mean changes in CV‐related Systemic Lupus International Collaborating Clinics/American College of Rheumatology Damage Index (SDI) from baseline to year 4 were estimated using weighted linear‐regression models. Time to first CV event up to four years was assessed using Kaplan–Meier estimators and a Cox regression model.

Results

Patients receiving anifrolumab accrued, on average, 0.046 fewer CV‐SDI points at four years (95% confidence interval [CI] −0.074 to −0.018; P = 0.008) compared with patients with real‐world SoC. The hazard ratio for CV events over four years was 0.257 (95% CI 0.0687 to 0.961; P = 0.0583) for patients treated with anifrolumab versus controls with real‐world SoC.

Conclusion

These findings suggest that anifrolumab may reduce long‐term CV damage in patients with SLE.

INTRODUCTION

Systemic lupus erythematosus (SLE) is a chronic autoimmune disease characterized by flares, progressive organ damage, and increased mortality. ¹ Current management strategies aim at remission or low disease activity, minimizing treatment‐related toxicity and comorbidities. ² Standard of care (SoC) in SLE includes antimalarials, nonsteroidal anti‐inflammatory drugs, glucocorticoids, and immunosuppressants. ² , ³ Recent international guidelines define biologic therapies as part of SoC, with clinical adoption beginning to expand. ³ Although SoC has improved clinical outcomes, patients with SLE still face a higher mortality risk compared with the general population. ⁴

Cardiovascular (CV) disease contributes significantly to this increased morbidity and mortality. ⁵ , ⁶ CV complications—including ischemic heart disease, pericarditis, and vascular disease—are particularly prevalent among patients with uncontrolled disease activity, recurrent flares, or prolonged glucocorticoid use. ⁷ , ⁸ , ⁹ Identifying effective therapies that reduce CV damage accrual is therefore essential to improving long‐term outcomes in SLE.

Anifrolumab, a human IgG1‐kappa monoclonal antibody that inhibits type I interferons, is a first‐in‐class biologic therapy approved for moderate to severe SLE in several countries. ¹⁰ , ¹¹ , ¹² , ¹³ The phase 3 randomized, double‐blind, placebo‐controlled Treatment of Uncontrolled Lupus via the Interferon Pathway (TULIP) trials demonstrated the efficacy of anifrolumab in managing disease activity while simultaneously achieving successful steroid tapering (TULIP‐1, NCT02446912; TULIP‐2, NCT02446899). ¹⁴ , ¹⁵ Long‐term safety and efficacy of anifrolumab were confirmed in a three‐year long‐term extension (LTE) of these trials (NCT02794285), although high dropout rates in the placebo arm and resulting imbalances between treatment cohorts limited robust assessment of long‐term organ‐damage outcomes. ¹⁶ , ¹⁷

To address this, the long‐term effectiveness of anifrolumab on organ damage was evaluated using an external control arm derived from real‐world (RW) data from the University of Toronto Lupus Clinic (UTLC) registry. ¹⁸ In the Long‐Term Organ Damage: Anifrolumab Versus Real‐World Standard of Care in Adult Patients With Active Systemic Lupus Erythematosus (LASER) study (NCT06485674), anifrolumab 300 mg plus SoC was associated with delayed and reduced overall organ damage over 208 weeks in patients with SLE compared with RW controls. ¹⁷ Given the importance of CV damage as a predictor of mortality, we built upon previous work ¹⁷ by focusing on CV‐specific damage accrual, a component of overall organ damage not previously investigated independently. Here, we present additional findings from the LASER study, focusing on the long‐term effectiveness of anifrolumab in reducing long‐term CV damage in adult patients with SLE, using RW external controls from the UTLC registry.

MATERIALS AND METHODS

Study design

This retrospective study analyzed clinical trial and observational data collected between January 1, 1995, and December 31, 2023. Patient consent was not required due to secondary use of anonymized data. CV‐specific damage outcomes in patients who initiated anifrolumab 300 mg while receiving SoC treatment in TULIP‐1 and TULIP‐2 (cohort of patients who received anifrolumab) were compared with external controls in the UTLC registry receiving RW SoC (RW SoC cohort). Patients in the cohort who received anifrolumab were indexed at the date of initiating anifrolumab and observed until death, loss to follow‐up, trial disenrollment, or week 208 assessment in the LTE study. Patients in the RW SoC cohort were indexed at the first clinical assessment date within the patient enrollment period for which they met all eligibility criteria and were receiving at least one eligible SoC treatment (ie, their “index assessment”). Follow‐up continued until death, loss to follow‐up, UTLC disenrollment, or end of study period.

Data sources

Data for the cohort of patients who received anifrolumab were obtained from the TULIP trials (TULIP‐1, TULIP‐2, and LTE). ¹⁴ , ¹⁵ , ¹⁶ TULIP‐1 (June 2015 to July 2018) evaluated the efficacy and safety of two doses of anifrolumab (150 and 300 mg every four weeks) versus placebo, whereas TULIP‐2 (July 2015 to September 2018) evaluated a single dose of anifrolumab (300 mg every four weeks) versus placebo. The LTE study (June 2016 to December 2021) allowed patients from TULIP‐1 and TULIP‐2 to continue receiving anifrolumab 300 mg or placebo for up to three additional years.

Data for the RW SoC external control cohort were obtained from the UTLC registry (January 1, 1995, to December 31, 2023). Established in 1970 in Toronto, Canada, the UTLC has been enrolling patients with SLE as part of an ongoing cohort study. Prospective data collection on organ damage, including CV damage, has been conducted annually using the Systemic Lupus International Collaborating Clinics/American College of Rheumatology Damage Index (SDI) ¹⁹ since 1995. Patients were evaluated every four weeks (cohort of patients who received anifrolumab) or every three to four months (RW SoC cohort) for demographic, clinical, laboratory, and treatment data, including variables identified as likely confounders for the present study.

Study population

All patients from TULIP‐1 and TULIP‐2 who initiated anifrolumab 300 mg were included in the cohort of patients who received anifrolumab. The RW SoC cohort comprised patients active in the UTLC database during the study period meeting key TULIP‐1 and TULIP‐2 eligibility criteria: age ≥18 years at index date, weight ≥40.0 kg, SLE diagnosed ≥24 weeks before index date using ≥4 of the 11 modified American College of Rheumatology classification criteria, ²⁰ , ²¹ a Systemic Lupus Erythematosus Disease Activity Index 2000 (SLEDAI‐2K) ≥6 points, no record of current pregnancy, and valid SDI measurement. Exclusion criteria at index date included receiving a glucocorticoid dose of >40 mg/day or any record of the following: (1) receiving any biologic agent at or within four weeks before the index date; (2) prior malignancy (except skin malignancy at least one year before index); (3) persistent, new, or recurrent nephrotic syndrome, chronic dialysis, or renal transplant; or (4) serum creatinine >2.0 mg/dL.

Drug exposures

Primary drug exposure was anifrolumab 300 mg, delivered intravenously every four weeks to patients in the TULIP trials, combined with SoC treatment. Glucocorticoid tapering to ≤7.5 mg/day between weeks 8 and 40 was attempted for those with a baseline dose of ≥10 mg/day in the cohort of patients who received anifrolumab. Comparison drug exposure consisted of RW SoC delivered according to standard clinical practice in the UTLC, without dosage or treatment‐length restrictions. Eligible SoC treatments in both cohorts included glucocorticoids, antimalarials, and immunosuppressants.

Study outcomes

The primary endpoint of this extended work was the change in CV‐specific SDI score at 208 weeks postindex date. The secondary study endpoint was time (in months) from index date until the first recorded increase in CV‐specific SDI scores, comprising angina or coronary artery bypass, myocardial infarction, cardiomyopathy or ventricular dysfunction, vascular disease, and pericarditis (at least six months) or pericardiectomy. CV outcomes were ascertained using identical SDI and CV‐specific definitions in both cohorts.

Baseline and confounding variables

Operational definitions of variables were standardized across the TULIP trials and the UTLC. In the TULIP trials, baseline characteristics were recorded at randomization or during a screening visit within 30 days before randomization. For patients in the UTLC registry, baseline data were collected at the initial assessment upon enrollment or during index assessment.

Baseline confounders were identified through consultation with clinical experts and review of similar published studies. Confounding variables (at index date) included age, SLE duration, sex, race, SLEDAI‐2K score, SDI score at baseline, proteinuria, glucocorticoid use and dose, antimalarial use, immunosuppressant use, hypertension, and smoking history. Both CV‐ and non‐CV–specific SDI scores were also included as confounding variables. Additional variables considered were year of diagnosis, age at diagnosis, year of index date, body mass index at index date, and serum creatinine level at index date.

Statistical analysis

Baseline characteristics were stratified by the treatment cohort and described as mean and SD or as frequency counts (n) and percentages (%). Missing baseline confounder data were imputed using a two‐stage approach: direct imputation using values recorded within three or more months of the index date if available, followed by multiple imputation.

Comparative analyses estimated the relative effect of anifrolumab 300 mg plus SoC in the TULIP trials compared with RW SoC in the UTLC cohort. Average treatment effect in the control (ATC) was selected as the primary estimand to ensure applicability to patients with RW SoC. This analysis used standardized mortality ratio weighting (SMRW) based on propensity scores to control baseline confounding. Standardized mean differences (SMDs) were used to assess baseline imbalance before and after SMRW. Inverse probability of censoring weighting (IPCW) was applied to account for the impact of informative loss to follow‐up.

The primary endpoint (change in CV‐specific SDI at 208 weeks) was estimated using a weighted linear‐regression model. Variables with residual imbalance post‐SMRW (SMD ≥ 0.2) were included as independent covariates in the regression model. Patient‐level treatment and censoring weights were calculated using 14 CV‐specific baseline confounders (Figure 1). Sensitivity analysis was performed to target both the average treatment effect in the treated (ATT; estimated through SMRW) and average treatment effect in the overlap (ATO; estimated through propensity score matching [PSM]) estimands. Robust sandwich estimators were used to calculate confidence intervals (CIs) for all estimated treatment effects.

Figure 1.

Balance of baseline confounding variables between treatments pre‐ and post‐SMRW (average treatment effect in the control population estimand). Standardized mean differences <0.2 (absolute value) indicate good covariate balance. SLEDAI‐2K score and proteinuria were included in outcome models to further improve balance. The values were obtained following imputation for missing variables. CV, cardiovascular; SDI, Systemic Lupus International Collaborating Clinics/American College of Rheumatology Damage Index; SLE, systemic lupus erythematosus; SLEDAI‐2K, Systemic Lupus Erythematosus Disease Activity Index 2000; SMRW, standardized mortality ratio weighting.

The secondary endpoint (time to first CV‐specific damage progression) was estimated using a Cox regression model. Censoring events included death, loss to follow‐up, TULIP or UTLC disenrollment, and end of study period. A weighted Kaplan–Meier plot was used to depict survival functions by the treatment cohort, and median survival time with 95% CI was reported. Hazard ratios (HRs) with 95% CIs were also reported from the Cox regression, with statistical significance set at P <0.05.

All statistical analyses were conducted using R, version 4.1.2. All data generated or analyzed during this study, which support the findings of this study, are included within this article and its supplementary files. Researchers interested in further analysis not presented in the article may contact the corresponding author. This retrospective study analyzed clinical trial and registry data precluding patient identification. Consequently, ethical approval was not required because this study involved secondary use of anonymized data.

RESULTS

The study population included 354 patients in the cohort who received anifrolumab and 561 patients in the RW SoC cohort; 175 patients (49.4%) and 345 patients (61.5%), respectively, completed 208 weeks of follow‐up.

Baseline characteristics are summarized in Supplementary Table S1. Most patients were female (anifrolumab, 92.4%; RW SoC, 88.4%), with a mean (±SD) age at index of 42.6 (±12.0) years and 34.1 (±12.2) years, respectively. The proportion of Black patients was lower in the cohort who received anifrolumab (13.0%) compared with the RW SoC cohort (21.6%), whereas White patients comprised 64.7% and 48.1% of each cohort, respectively. Mean ± SD disease duration was longer in the cohort of patients who received anifrolumab (9.6 ± 8.6 years) compared with RW SoC controls (6.8 ± 6.5 years). SLEDAI‐2K scores (mean ± SD) were similar (anifrolumab, 11.4 ± 3.8; RW SoC, 10.4 ± 4.9), with approximately one‐third of the patients having an SDI score ≥1 in both treatment cohorts (32.0% vs 36.2%, respectively). Baseline glucocorticoid use and mean (±SD) daily dose were higher in patients with RW SoC (96.4%; 14.7 ± 10.2 mg/day) than in the cohort of patients who received anifrolumab (81.4%; 9.0 ± 7.3 mg/day). Proteinuria was more common in the RW SoC cohort (39.6%) than in the patients treated with anifrolumab (4.8%), whereas rates of hypertension (RW SoC, 29.8%; anifrolumab, 32.2%) and smoking history (RW SoC, 21.4%; anifrolumab, 22.3%) were comparable between cohorts.

To account for baseline confounding and potential information loss to follow‐up, patient characteristics were adequately balanced using PSM and IPCW, as previously described. ¹⁷ Following SMRW (ATC estimand), adequate balance was achieved across treatment cohorts, except for SLEDAI‐2K score (SMD −0.24) and proteinuria (SMD −0.20), which were included as independent covariates in the outcome model (Figure 1). Adequate balance was also achieved for all confounding variables when targeting the ATT estimand (Supplementary Table S2).

Mean change in CV‐specific SDI from index to week 208 was 0.046 points lower (95% CI −0.074 to −0.018; P = 0.008) in the weighted cohort of patients who received anifrolumab than in the RW SoC cohort (ATC; Table 1). Sensitivity analysis considering the ATT (–0.063; 95% CI –0.119 to –0.007; P = 0.001) was consistent with ATC results. Following PSM in the subgroup of patients with ≥208 weeks of follow‐up, no significant change in CV‐specific SDI was observed between treatment cohorts (ATO –0.038; 95% CI –0.086 to 0.010; P = 0.161; Table 1).

Table 1.

Estimated difference between treatment cohorts in cardiovascular‐specific SDI score between index and 208 weeks postindex (all estimands)^*

Estimand	Type of regression analysis	Method	Parameter	TULIP cohort	UTLC cohort	Mean difference of change in SDI between cohorts	95% CI	Robust SE	P value
Main analysis
ATC	Linear regression a	SMRW (untreated) plus IPCW b	N	354	561	−0.046	−0.074 to −0.018	0.014	0.008
			Mean change in SDI	0.005	0.046
			Robust SE	0.004	0.013
			95% CI	−0.002 to 0.013	0.021 to 0.071
Sensitivity analysis
ATT	Linear regression	SMRW (treated) plus IPCW c	N	354	561	−0.063	−0.119 to −0.007	0.029	0.001
			Mean change in SDI	0.013	0.076
			Robust SE	0.009	0.027
			95% CI	−0.005 to 0.031	0.023 to 0.129
ATO	Linear regression	PSM d	N	110	110	−0.038	−0.086 to 0.010	0.024	0.161
			Mean change in SDI	0.017	0.056
			Robust SE	0.012	0.021
			95% CI	−0.007 to 0.041	0.014 to 0.097

^*To account for the impact of informative censoring, IPCW was used. Censoring weights were calculated from the ratio of stabilized censoring probability to censoring probability. ATC, average treatment effect in the control population; ATO, average treatment effect in the overlap population; ATT, average treatment effect in the treated population; CI, confidence interval; IPCW, inverse probability of censoring weighting; PSM, propensity score matching; SDI, Systemic Lupus International Collaborating Clinics/American College of Rheumatology Damage Index; SMRW, standardized mortality ratio weighting; TULIP, Treatment of Uncontrolled Lupus via the Interferon Pathway; UTLC, University of Toronto Lupus Clinic.

^a The linear‐regression model for SMRW (untreated) plus IPCW was further adjusted for Systemic Lupus Erythematosus Disease Activity Index 2000 score and proteinuria level at the index date because these variables remained unbalanced after weighting (ie, standardized mean difference > 0.2).

^b Treatment weights were calculated using the SMRW method, with the UTLC cohort as the target population (ATC) for the main analysis (SMRW [untreated] plus IPCW).

^c Treatment weights were calculated using the SMRW method, with the TULIP cohort as the target population (ATT) for the sensitivity analysis (SMRW [treated] plus IPCW).

^d The PSM method was applied to estimate ATO in patients with ≥208 weeks of follow‐up. Nearest‐neighbor matching with replacement was applied to estimate propensity scores within a caliper value defined as 20% of SD for the distribution of the propensity score variable in the full sample.

A total of 6 patients in the cohort who received anifrolumab experienced CV events compared with 17 patients in the RW SoC cohort (Supplementary Table S3). The HR for CV events over four years was 0.257 (95% CI 0.0687–0.961; P = 0.0583; Figure 2) for anifrolumab versus RW SoC. A modest separation of the weighted Kaplan–Meier curves was observed, with an estimated 99% of the patients who received anifrolumab (95% CI 98–100%) and 96% of those received RW SoC (95% CI 95–98%) not experiencing CV damage progression by 208 weeks.

Figure 2.

Weighted KM curves for time to cardiovascular damage progression by treatment cohort up to four years, with estimated hazard rates from fitted Cox models (average treatment effect in the control population estimand). HR and hazard rates refer to the effect estimates by the Cox model, with the hazard rate in each cohort averaged across the observed period. The Cox regression model for SMRW (untreated) plus IPCW was further adjusted for Systemic Lupus Erythematosus Disease Activity Index 2000 score and proteinuria level at the index date because these variables remained unbalanced after weighting (ie, standardized mean difference > 0.2). Restricted mean survival time and minimum and maximum survival times were derived while accounting for the combined SMRW plus IPCW weights (ie, the raw survival times were weighted). Hence, some extreme weighted values are beyond the 48 months of observed time. CI, confidence interval; HR, hazard ratio; IPCW, inverse probability of censoring weighting; KM, Kaplan–Meier; RW SoC, real‐world standard of care; SMRW, standardized mortality ratio weighting.

DISCUSSION

Although SLE‐related mortality has reduced considerably in recent years, CV damage remains one of the leading causes of death in patients with SLE. Observational studies have found that patients with SLE—particularly premenopausal women—face a significantly increased risk for CV events, with estimates ranging from 2‐ to 50‐fold higher than age‐matched controls. ⁶ , ²² , ²³ This irreversible damage accrual in SLE has been linked to both disease activity and treatment‐related factors. ²³ , ²⁴ , ²⁵ Higher SLEDAI‐2K scores are associated with vascular arterial complications, including ischemic heart disease, peripheral artery disease, and cerebrovascular events. ²⁴ , ²⁶ Prolonged or high‐dose glucocorticoid use is also independently associated with increased CV risk through hypertension, dyslipidemia, and insulin resistance. ⁷ , ⁸ Treatment strategies that control disease activity while minimizing glucocorticoid intake are therefore warranted to reduce long‐term CV damage in patients with SLE, as recommended by EULAR. ²⁷

A recent analysis of TULIP trial data compared with RW external controls showed that reducing disease activity and tapering glucocorticoid use with anifrolumab was associated with lower rates of organ‐damage accrual in patients with SLE. ¹⁷ Building on this, we assessed the effect of anifrolumab plus SoC compared with RW SoC on long‐term CV damage accrual. After baseline adjustment, our results indicate that CV damage in patients treated with anifrolumab plus SoC in the TULIP trials was lower than in the RW SoC cohort over a four‐year period. This difference in CV damage accrual may be driven by reduced disease activity, lower glucocorticoid use, or potential anifrolumab‐specific anti‐inflammatory vascular effects, although the underlying mechanism remains unclear. ²⁸ , ²⁹ Given that CV damage gradually accumulates over 10 to 15 years, the observed reduction in CV‐SDI with anifrolumab over four years is numerically small but directionally consistent with a potential long‐term cardioprotective effect, supporting the need for longer follow‐up. ¹⁹ , ³⁰

To minimize bias, we applied a target trial emulation framework, a robust method typically recommended for analyzing external control data. ¹⁷ We mirrored key eligibility criteria from patients in the TULIP trials to reduce selection bias, used propensity score weighting to address baseline confounding while preserving sample size, and applied censoring weights to correct for loss to follow‐up. The study also leveraged a high‐quality RW cohort used in numerous other SLE studies. However, we recognize that although propensity score methods help address confounding in observational studies, they cannot replace randomized controlled trials. Study limitations include potential imbalance between treatment cohorts due to unaccounted time‐varying factors and residual differences in baseline characteristics potentially associated with renal involvement (eg, proteinuria, disease activity, and immunosuppressant use), as well as the risk for differential outcome misclassification when integrating trial and RW data. However, significant bias is unlikely given that SDI changes slowly over time and both cohorts had a similar annual recording frequency. The influence of evolving SoC, including the introduction of mycophenolate mofetil in the early 2000s and biologics since 2020, was addressed by excluding patients with recent biologic use (within four weeks of index), but some exposure during follow‐up may have occurred. Additionally, proactive glucocorticoid tapering in the anifrolumab group may have influenced outcomes, and residual confounding and time‐era differences in CV risk factor management (eg, hypertension and hyperlipidemia) cannot be fully ruled out. Overall, our findings are consistent with a potential benefit of anifrolumab versus RW SoC in reducing long‐term CV organ damage, highlighting the need for continued research into CV‐protective therapies for patients with SLE.

AUTHOR CONTRIBUTIONS

All authors contributed to at least one of the following manuscript preparation roles: conceptualization AND/OR methodology, software, investigation, formal analysis, data curation, visualization, and validation AND drafting or reviewing/editing the final draft. As corresponding author, Dr Chandran confirms that all authors have provided the final approval of the version to be published, and takes responsibility for the affirmations regarding article submission (eg, not under consideration by another journal), the integrity of the data presented, and the statements regarding compliance with institutional review board/Declaration of Helsinki requirements. All authors contributed to the investigation, formal analysis of the data, and writing, reviewing, and editing of the manuscript.

ROLE OF THE STUDY SPONSOR

Medical writing and editorial assistance were provided by Paula de Oliveira, Chrysi Petraki, and Daria Renshaw of IQVIA, funded by AstraZeneca. This study was sponsored by AstraZeneca.

Supporting information

Disclosure Form:

Table S1. Baseline characteristics by treatment cohort.

Table S2. Balance of baseline confounding variables between treatments pre‐ and post‐SMRW (ATT estimand).

Table S3. Description of time to first cardiovascular event from index date.