Longitudinal associations of flare and damage accrual in patients with systemic lupus erythematosus

Abstract

Objective

To estimate the prevalence of organ damage (damage) and flare and to examine longitudinal associations between flares and subsequent damage accrual, in patients with systemic lupus erythematosus (SLE).

Methods

Patients enrolled in the Asia Pacific Lupus Collaboration cohort with ≥3 years of prospectively captured data were studied. Flares were assessed at routine visits, while damage ((Systemic Lupus International Collaborating Clinics/American College of Rheumatology) Damage Index) was assessed annually. Multivariable, multifailure survival analyses were carried out to quantify the association between flares and damage accrual.

Results

1556 patients with SLE with a median (IQR) of 5.7 (3.9, 7.0) years of follow-up were studied. 39.5% (n=614) of patients had damage at enrolment, and 31.9% (n=496) accrued damage during the study observation period. The incidence of damage accrual during observation was ~58/1000 person-years. Overall, 74.1% (n=1153) of patients experienced a flare of any severity (mild/moderate or severe) at least once; 56.9% (n=885) experienced recurrent (≥2) flares. The risk of subsequent damage accrual in patients who experienced mild-to-moderate flare, after controlling for confounders, was 32% greater than in patients without flares (adjusted HR) (95% CI 1.32 (1.17 to 1.72)). The risk of damage accrual was greater if patients had severe flares (HR (95% CI) 1.58 (1.18 to 2.11)). For each additional flare, the risk of damage accrual increased by 7% (HR (95% CI) 1.07 (1.02 to 1.13)).

Conclusions

Flares independently increased the risk of damage accrual. Prevention of flares should be considered a necessary goal of SLE disease management to minimise permanent damage.

WHAT IS ALREADY KNOWN ON THIS TOPIC.

WHAT THIS STUDY ADDS

• Using a large, multicentre, multinational prospective SLE cohort, we provide evidence that the risk of damage accrual increased with increasing number of flares experienced by patients with SLE, especially if patients experienced severe flares.

HOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE OR POLICY

• Findings from this study further confirm that flares are independent risk factors for organ damage accrual in patients with SLE. Prevention of flares to minimise organ damage should be a treatment goal of SLE disease management.

Introduction

Systemic lupus erythematosus (SLE) is a chronic, heterogeneous autoimmune disease, affecting multiple organs, including skin, joints, kidney, heart, lungs and the central nervous system.¹ At least five million people worldwide are affected by SLE,² with approximately 90% of these individuals being women diagnosed between ages of 15 and 44,3,5 among whom SLE is one of the top 15 causes of death.⁶ The natural history of SLE is characterised by periods of disease activity interspersed with periods of relative latency.⁷ Exacerbations of disease activity, generally referred to as flares, vary in severity from mild-to-moderate episodes that can be managed in outpatient clinic settings, to severe episodes necessitating hospital admission.⁷ Such flares have been reported to be associated with irreversible organ damage (damage) and premature death.^{8 9}

Treatment for SLE has seen minimal changes over the past half-century.³ Despite the approval of belimumab in 2011^{10 11} and anifrolumab in 2021,^{12 13} approximately 80% of patients are treated with glucocorticoids (GC), resulting in harmful adverse effects, including irreversible damage, independent of SLE disease activity.^{14 15} Approximately 30% to 50% of patients with SLE develop damage within 5 years of diagnosis, predominantly due to uncontrolled disease activity and treatment toxicity.16,18 Limiting the rate and severity of flares in SLE is a key goal, with extensive research focused on the development of flare prediction biomarkers and flare risk.¹⁹ Moreover, the EULAR 2023 recommendations suggest that the minimisation of GC use is a central goal in SLE management.²⁰

Accurate data quantifying the association between flare and damage are sparse, as are high-quality data on the incidence of damage accrual in patients with SLE, including in Asian countries,²¹ predominantly due to lack of prospective studies focused on these aspects. We sought to investigate the burden of flares and damage in SLE using the Asia Pacific Lupus Collaboration (APLC) patient cohort (NCT03138941), which is an ongoing prospective study of patients with SLE recruited from multiple countries in the Asia Pacific region.²² Furthermore, we used the large size of this dataset to quantify the independent association between flares and subsequent damage accrual.

Methods

Patients

In this study (GlaxoSmithKline (GSK) study 217856), patients with SLE enrolled in the APLC cohort with a minimum of 3 years of observational data without exposure to biologics (ie, rituximab or belimumab), captured prospectively between 2013 and 2020 and were studied (figure 1A). Study participants were recruited from centres in Australia, China, Indonesia, Japan, Korea, Malaysia, the Philippines, Singapore, Taiwan and Thailand and received standard of care treatment. All APLC patients were consenting adults who met either the 1997 American College of Rheumatology (ACR) Modified Classification Criteria for SLE²³ or the Systemic Lupus International Collaborating Clinics (SLICC) 2012 Classification Criteria.¹⁶

Figure 1. Study cohort, number (%) flares, damage and damage accrual. (A) Selection criteria of the study cohort; (B) proportion of patients with flares and organ damage at baseline visit and by last visit; (C) proportion of patients with damage accrual by number of total flares. ACR, American College of Rheumatology; APLC, Asia Pacific Lupus Collaboration; BEL, belimumab; damage, organ damage; RTX, rituximab; SDI, SLICC/ACR Damage Index; SLE, systemic lupus erythematosus; SLICC, Systemic Lupus International Collaborating Clinics.

Patient and public involvement

Neither patients nor the public were involved in the design or conduct of this study.

Variables

The APLC cohort prospectively captures patient demographics including gender, ethnicity and years of birth, SLE onset and SLE diagnosis, and classification criteria (ACR²³ and SLICC)¹⁶ at recruitment. Damage was assessed using the SLICC/ACR Damage Index (SDI),²⁴ measured at recruitment and serially at annual visits. Damage was defined as present if the SDI was greater than zero (SDI >0), while damage accrual was defined if the change in SDI from baseline was greater than zero (ΔSDI >0). Data on disease activity, medications and pathology results were captured at each 3–6 monthly routine visit. Disease activity was measured using the SLE Disease Activity Index (SLEDAI)—2000 (2K)^{25 26} and Physician Global Assessment of activity (PGA: 0–3).²⁷ Presence of mild-to-moderate or severe flares was assessed using the Safety of Estrogens in Lupus Erythematosus National Assessment (SELENA)-SLEDAI flare index (SFI).^{28 29}

Statistical analysis

Statistical analysis was performed using Stata V.18 (StataCorp, College Station, Texas, USA). Since the SDI data were only captured at annual visits, damage data at other visits without SDI scoring were replaced by carrying over the nearest known SDI values using either carry-backward or carry-forward methods. This was performed based on the assumption that damage in SLE is irreversible and therefore, SDI scores cannot decrease. The distribution of SDI data was assessed using probability plots, including Quantile-Quantile plots and cumulative distribution function plots. The data were observed to be skewed, as indicated by deviations from the straight diagonal line and the presence of an upward curvature (online supplemental figure 1). Organ damage accrual was hence analysed as a binary variable, categorised as either new damage or no new damage during the study period.

Patient characteristics are described as summary statistics. Continuous variables are expressed as median (IQR, range) and compared using Wilcoxon rank sum tests. Categorical variables are described as frequency (%) and compared using χ² tests. The longitudinal associations between flares and subsequent damage accrual were examined using multifailure survival analyses incorporating Prentice, Williams and Peterson—total time (PWP-TT) models,^{30 31} accounting for any potential confounding effects. Clustering was specified in the Cox regression models to account for intragroup correlation. We first examined the impact of flare regardless of the severity (ie, any flare, mild/moderate or severe), followed by flares stratified by severity (ie, mild/moderate and severe flares separately compared with no flare). We treated flares as categorical as well as numerical variables (cumulative number of flares). Overall, five multivariable regression models were performed with different definitions of flares (at previous visit (_t-1)) as the primary exposure variable and damage accrual (at visit_t) as the outcome (in all models):

$${\displaystyle 1.Anyflare{\left(M/M/S\right)}_{t-1}comparedtonoflare\left(binary\right)⟶Damageaccrua{l}_t}$$

$${\displaystyle \begin{array}{l}2.M/Mflar{e}_{t-1}\mathrm{\&}Sflar{e}_{t-1}exclusivelycomparedtonoflare\left(3cat.\right)\\ ⟶Damageaccrua{l}_t\end{array}}$$

$${\displaystyle 3.Cumulativenumberofany\left(M/M/S\right)flare{s}_{t-1}\left(count\right)⟶Damageaccrua{l}_t}$$

$${\displaystyle 4.CumulativenumberofM/Mflare{s}_{t-1}\left(count\right)⟶Damageaccrua{l}_t}$$

$${\displaystyle 5.CumulativenumberofSflare{s}_{t-1}\left(count\right)⟶Damageaccrua{l}_t}$$

*M/M/S=mild-to-moderate or severe

Time-invariant baseline characteristics (eg, gender, ethnicity, smoking status, education level, gross domestic product (GDP)) and time-dependent variables measured until the visit prior to damage accrual (eg, age and duration) were considered. If univariable associations of covariates with damage accrual produced associations with p<0.1, they were considered for inclusion in at least the first multivariable models. Retention of potential confounders in the models depended on their statistical significance and the impact on magnitude of the beta coefficient of primary exposure (ie, HR of flare). The results from regression analyses are presented as HRs with corresponding 95% CI. P values ≤0.05 are considered statistically significant.

Diagnostic tests for all models were performed to assess how well the models fit the data and are reported in online supplemental table 3. While all the statistical models were extensions of Cox regressions incorporating PWP-TT, all models met the proportional hazards assumption. Akaike Information Criterion (AIC) and Bayesian Information Criterion (BIC) were used to assess the relative quality of validity of the statistical models considering the analytical dataset. Lower AIC and BIC indicate better/improved statistical models.³² The second model (with three categorical flare variables) was the best fitted model. Kaplan-Meier survival curves were drawn and log-rank test was performed to test the equality of survivor functions of mild-to-moderate or severe flares compared with no flare.

Ethics approval

This study was sponsored by GSK (study ID 217856) and conducted by Monash University on behalf of the APLC in accordance with the National Statement of Ethical Conduct in Human Research 2023.³³ Each APLC site has local ethics approval for patient recruitment and to contribute to the centralised APLC dataset. Individual centres obtain written informed consent in accordance with local authorities regarding ethical conduct of human research. Monash University Human Research Ethics Committee has approved to store the central dataset in Monash University’s secure servers and to perform analyses using collective data (MUHREC Project ID 18778).

Results

Patient characteristics

A total of 1556 patients with ≥3 years of follow-up data were studied. All patients had ≥2 SDI surveys completed. Study participants were followed up for a median (IQR) of 5.7 (3.9, 7.0) years, for a total of 26 164 visits. In total, 93.4% of study participants were female and the median (IQR) age at study enrolment into the APLC registry and diagnosis was 40 years (31, 51) and 30 years (22, 40.5), respectively (table 1). The majority of patients were of Asian ethnicity (89.1%) and 15.7% were enrolled within ≤1 year from SLE diagnosis. Based on GDP purchasing power parity per capita in 2020, Australia, Singapore and Taiwan were grouped as countries with GDP ≥international dollars (Int$) 50 000; Japan, Korea and Malaysia were grouped in ‘<Int$50 000 and ≥Int$20 000’ category; and China, Indonesia, Philippines and Thailand were grouped in GDP <Int$20 000 category. Approximately 39% of the primary cohort was from countries with GDP <Int$20 000 (table 1).

Table 1. Patient characteristics.

	Study cohort
	n=1556 n (%) or median(IQR)
Demographics
Age at enrolment (years)	40 (31, 51)
Age at diagnosis (years)	30 (22, 40.5)
Disease duration at enrolment (years)	8 (3, 14)
Study duration (years)	5.7 (3.9, 7.0)
Inception cohort*	244 (15.7%)
Females	1453 (93.4%)
Asian ethnicity†	1386 (89.1%)
Current smoker at enrolment‡	76 (4.9%)
SLE family history§	115 (7.4%)
Education level¶
Primary	236 (15.2%)
Secondary	609 (39.3%)
Tertiary	704 (45.4%)
GDP (PPP) per capita (Int$)
≥50 000	758 (48.7%)
<50 000 and ≥20 000	198 (12.7%)
<20 000	600 (38.6%)
Medication use at enrolment (baseline)
GC use	1215 (78.1%)
Daily GC dose (mg/d)	5.0 (2.0, 10.0)
AM use	1084 (69.7%)
IS use	814 (52.3%)
Clinical indicators at enrolment
SLEDAI-2K score	4.0 (2.0, 6.0)
PGA	0.5 (0.2, 1.0)
In LLDAS	715 (46.2%)
Any flare (M/M/S) present	222 (14.3%)
M/M flare present	174 (11.2%)
Severe flare present	81 (5.2%)
Damage present	614 (39.5%)
SDI score	0.0 (0.0, 1.0)
Medication use during study observation period
GC-Ever (at least once)	1344 (86.4%)
TAM-GC (mg/d)	4.9 (2.0, 7.6)
Cumulative GC dose (g)	9.2 (3.5, 14.7)
AM-Ever	1217 (78.2%)
IS-Ever	1134 (72.9%)
Clinical indicators during study observation period
LLDAS achieved at least once (Ever)	1403 (90.2%)
Cumulative %time in LLDAS	50.4 (21.3, 75.3)
TAM-SLEDAI-2K (AMS)	3.0 (1.5, 4.7)
TAM-PGA	0.4 (0.2, 0.7)
Any (M/M/S) flare—Ever	1153 (74.1%)
Number of any flares	2.0 (0.0, 4.0)
M/M flare—Ever	1120 (72.0%)
Number of M/M flares	2.0 (0.0, 3.0)
Annual M/M flare rate	0.3 (0.0, 0.6)
Severe flare—Ever	599 (38.5%)
Number of severe flares	0.0 (0.0, 1.0)
Annual severe flare rate	0.0 (0.0, 0.3)
Damage present at last visit	863 (55.5%)
SDI score at last visit	1.0 (0.0, 2.0)
Damage accrued during study period	496 (31.9%)
Deceased patients	24 (1.5%)

^*Patients who were ≤1 year from SLE diagnosis.

^†N=1555.

^‡N=1551.

^§N=1552.

^¶N=1449.

AM, antimalarial medications; AMS, adjusted mean SLEDAI; damage, organ damage; GC, glucocorticoid; GDP, gross domestic product; Int$, international dollars; IS, immunosuppressant; LLDAS, Lupus Low Disease Activity State; M/M/S, mild/moderate/severe; PGA, Physician Global Assessment; PPP, purchasing power parity; SDI, Systemic Lupus International Collaborating Clinics/American College of Rheumatology Damage Index; SLE, systemic lupus erythematosus; SLEDAI-2K, SLE Disease Activity Index–2000; TAM, time adjusted mean

At enrolment, 78.1% of patients were on GC, 69.7% on antimalarial medications (AM) and 52.3% were on immunosuppressants (IS). During the study observation period, 86.4% of patients used GC, 78.2% used AM and 72.9% used IS medications at least once (ever users). In terms of disease activity indicators, patients’ median (IQR) SLEDAI-2K and PGA at enrolment were 4 (2, 6) and 0.5 (0.2, 1.0), respectively. The median (IQR) time-adjusted mean (TAM) SLEDAI-2K during the study observation period (also referred to as AMS)³⁴ was 3.0 (1.5, 4.7) while the median (IQR) TAM PGA was 0.4 (0.2, 0.7).

Burden of damage

Overall, 614/1556 patients had irreversible damage present at enrolment; hence, the prevalence of damage at enrolment was 39.5% (95% CI 37.0% to 41.9%). Likewise, the prevalence of damage at last study observation visit was 55.5% (95% CI 53.0% to 58.0%). Approximately 32% of the study cohort, including those with damage at baseline, accrued damage during the study observation period. Thus, the estimated incidence of damage accrual of the study cohort was ~58/1000 person-years (95% CI 53/1000 to 63/1000). Median (IQR) number of days to first new damage was 879.5 (525, 1428) (~2.4 (1.4, 3.9) years). A total of 693 patients remained damage-free at the end of the study observation period.

Damage accrued in all 12 domains assessed using the SDI: ocular, neuropsychiatric, renal, pulmonary, cardiovascular, peripheral vascular, gastrointestinal, musculoskeletal, skin, premature gonadal failure, diabetes and malignancy. Musculoskeletal damage was the most prevalent, followed by ocular, neuropsychiatric and renal damage at the first visit, with gastrointestinal damage as the least prevalent at last visit (online supplemental figure 2).

Patients who had damage at enrolment were significantly older and had a longer disease duration. Proportions of patients with non-Asian ethnicity were significantly higher in patients with baseline damage, compared with patients without (online supplemental table 1). Higher proportions of patients with baseline damage were on GC (81.9% vs 75.6%, p=0.003) and IS (56.4% vs 49.7%, p=0.010) at study enrolment; in contrast, the proportion of patients on AM was significantly lower (62.2% vs 74.5%, p<0.001). In terms of disease activity at enrolment, SLEDAI-2K and flares were similar between patients with or without baseline damage. These trends in medication use, SLEDAI-2K and flares persisted during the study observation period (online supplemental table 1).

A significantly higher proportion of patients with baseline damage accrued further damage (40.2% vs 26.4%, p<0.001; online supplemental table 1) compared with patients without baseline damage. Patients who accrued damage during the study period were older with longer disease, and study durations and comparably fewer patients had tertiary education compared with patients who did not accrue further damage (online supplemental table 1). Significantly higher proportions of patients with damage accrual were smokers at enrolment and from countries with GDP <int$20 000. The use of AM at enrolment and during the study period was significantly lower in patients who accrued damage. In contrast, the use of GC or IS at least once during the observation period was significantly higher. TAM SLEDAI-2K, TAM PGA and the proportion of patients who experienced flares were significantly higher in patients who accrued damage.

Burden of flares

Approximately 14.3% of patients were classified as having a flare at enrolment (table 1 and figure 1B); 11.2% had mild-to-moderate flares and 5.2% had severe flares. Nearly three-quarters of the study cohort (74.1% (n=1153/1556)) experienced ≥1 flare (mild/moderate or severe) and 56.9% (n=885/1556) experienced recurrent (≥2) flares during the study period. Of the patients who experienced flares (flare-ever), 34.3% (395/1153) accrued new damage compared with 25.1% (101/403) of patients who never flared (p<0.001). More patients who had recurrent (≥2) flares accrued damage (figure 1C).

Longitudinal associations between flares and damage accrual

We next examined the impact of flare on subsequent damage accrual using univariable and multivariable, multifailure survival analyses. We observed a statistically significant association between flares and subsequent damage accrual. Based on univariable survival models, if a patient had a mild-to-moderate flare (excluding severe flare) at any given visit during the study period, the probability of the patient accruing new damage at any given subsequent visit was 38% greater (unadjusted HR=1.38 (95% CI 1.08 to 1.78), p=0.011), compared with those without flares (table 2 and figure 2). The risk of damage accrual was doubled in patients who experienced severe flares (unadjusted HR=2.03 (95% CI 1.56 to 2.65), p<0.001). The number of flares was also associated with increased risk of damage accrual, whereby if the cumulative number of any flares increased by one, we observed the risk of damage accrual at a given subsequent visit to increase by 14% (unadjusted HR=1.14 (95% CI 1.09 to 1.19), p<0.001) (table 2).

Table 2. Univariable associations of damage accrual (ΔSDI>0) at t_n.

	HR	(95% CI)	P value
Demographics
Age at tn−1 (years)	1.02	(1.02 to 1.03)	<0.001
Disease duration at tn−1 (years)	1.02	(1.01 to 1.03)	<0.001
Male gender	0.94	(0.68 to 1.29)	0.7
Asian ethnicity	1.06	(0.81 to 1.40)	0.7
Smoker at enrolment	1.21	(0.90 to 1.63)	0.2
Tertiary education	0.72	(0.60 to 0.85)	<0.001
GDP Int$<20 000	1.54	(1.27 to 1.86)	<0.001
Medication use at tn−1
GC use	1.64	(1.34 to 2.01)	<0.001
Cumulative GC dose (g)	1.04	(1.03 to 1.05)	<0.001
AM use	0.69	(0.58 to 0.81)	0.003
IS use	1.28	(1.08 to 1.52)	0.005
Disease activity/damage at tn−1
Time adjusted mean SLEDAI-2K (AMS)	1.10	(1.07 to 1.14)	<0.001
Damage (any) present	1.65	(1.40 to 1.95)	<0.001
Flares at tn−1
Flare absent	1.00
Any (M/M/S) flare present	1.60	(1.33 to 1.94)	<0.001
Flare absent	1.00
M/M flare present	1.38	(1.08 to 1.78)	0.011
Severe flare present	2.03	(1.56 to 2.65)	<0.001
Cumulative number of flares at tn−1
Any (M/M/S) flares	1.14	(1.09 to 1.19)	<0.001
M/M flares	1.14	(1.08 to 1.20)	<0.001
Severe flares	1.17	(1.11 to 1.24)	<0.001

AMantimalarial medicationsAMStime adjusted mean SLEDAI-2KGCglucocorticoidsGDPgross domestic productISimmunosuppressantsM/M/Smild/moderate/severeSDISystemic Lupus International Collaborating Clinics/American College of Rheumatology Damage IndexSLEDAI-2KSLE Disease Activity Index—2000t_n-1, previous visitt_n, any given visit

Figure 2. Kaplan-Meier graphs of survival probabilities of damage accrual. P value from log-rank test assessing the equality of survivor functions is <0.001. M/M, mild/moderate.

In addition to flares, higher TAM SLEDAI-2K (AMS) and the presence of damage at enrolment were each significantly associated with damage accrual. In terms of demographic factors, older age, longer disease duration and low GDP were significant risk factors for damage accrual. Tertiary education was negatively associated with damage accrual in a univariable analysis, but this association was attenuated (ie, no longer statistically significant) when adjusted for confounding factors (online supplemental table 2). Likewise, AM use demonstrated a protective effect against damage accrual in a univariable analysis, but this association was no longer statistically significant in multivariable models. On the other hand, GC use was strongly associated with damage accrual, while IS use was associated with borderline statistical significance, after adjusting for other confounding factors (online supplemental table 2).

After adjusting for potential confounding factors, flares remained a significant predictor of damage accrual at subsequent visits. If a patient experienced a mild-to-moderate flare, the risk of damage accrual was increased by 32% (adjusted HR=1.32 (95% CI 1.03 to 1.70), p=0.029) and if it was a severe flare, the risk of subsequent damage accrual was 58% (adjusted HR=1.58 (95% CI 1.18 to 2.11), p=0.002) greater than the patients without severe flares (table 3 and online supplemental table 3). For each additional (any severity) flare, after controlling for confounders, the risk of damage accrual increased by 7% (adjusted HR=1.07 (95% CI 1.02 to 1.13), p=0.011) (table 3 and online supplemental table 3).

Table 3. Independent associations between flares and damage accrual (ΔSDI>0) at t_n.

	HR*	(95% CI)	P value
Flares at tn−1
Flare absent	1.00
Any (M/M/S) flare present	1.42	(1.17 to 1.72)	<0.001
Flare absent	1.00
M/M flare present	1.32	(1.03 to 1.70)	0.029
Severe flare present	1.58	(1.18 to 2.11)	0.002
Cumulative number of flares at tn−1
Any (M/M/S) flares	1.07	(1.02 to 1.13)	0.011
M/M flares	1.08	(1.01 to 1.14)	0.015
Severe flares	1.05	(0.98 to 1.13)	0.13

^*HRs adjusted for patient age at t_n−1, disease duration at t_n−1, gross domestic product of patients’ country, cumulative prednisolone dose at t_n−1, immunosuppressant use at t_n−1, time-adjusted SLEDAI score at t_n−1, and for the presence of existing organ damage at t_n−1. See See Supplementary Table 3online supplemental table 3 for details.

Damageorgan damageM/M/Smild/moderate/severeSDISystemic Lupus International Collaborating Clinics/American College of Rheumatology Damage IndexSLEDAISLE Disease Activity Indext_n-1, previous visitt_n, any given visit

Discussion

SLE is characterised by flares, progressive damage and increased mortality and predominantly occurs in women of childbearing age.^{5 35} Damage reflects irreversible changes occurring in patients with SLE as a result of inadequately controlled disease activity, SLE-associated comorbidities and treatment.³⁶ Quantitative enumeration of the association between flare and damage in SLE may inform the use of treatments designed to reduce flares, or add meaning to the results of clinical trials where such effects are observed. Such quantification requires the use of large carefully collected prospective datasets.

This study is a comprehensive examination using the real-world evidence of burden of flare and damage in a large prospectively followed cohort of patients with SLE. Reducing incident irreversible damage in SLE has been the central goal of care, as damage has been associated with mortality.^{15 20 37} The prevalence of damage in the study cohort was high; nearly 40% of study participants had damage at study entry and 55% had damage at the last visit. Patients in the study cohort had been living with SLE for an average of 8 years, from the time of SLE diagnosis until the study enrolment, a cohort characteristic that may contribute to the high prevalence of damage at study entry. About 58 patients per 1000 patient-years accrued new damage (~6% per year), with overall, one-third of the study cohort having damage accrual during the observation period of nearly 6 years. Patients who accrued new damage were older, had longer disease duration and had evidence of more severe disease; these are all well-known risk factors for damage accrual in patients with SLE.³⁶ Furthermore, significantly higher proportions of patients with damage accrual were current smokers and from lower GDP countries (<int$20 000). Smoking and low socioeconomic status are also known risk factors for damage accrual.^{38 39}

Limiting the frequency and severity of flares is a core objective in SLE disease management and forms part of the most recent guidelines for SLE management.²⁰ Despite this goal, nearly three quarters (74.1%) of the study cohort patients experienced a flare at least once during the study period. Compared with patients without damage accrual, a significantly higher proportion of patients with damage accrual had experienced flares during the study period. The presence of flare was a significant risk factor for subsequent damage accrual and the risk was higher if flares were severe or recurrent. The estimated increased risk of damage accrual at any subsequent visit was 32% for patients experiencing mild-to-moderate flares and 58% for patients experiencing severe flares. Increased numbers of flares, irrespective of their severity, also significantly increased the risk of subsequent damage accrual. A similar association between number of flares (of any severity) and increased risk of damage has been shown in the Grupo Latino Americano De Estudio del Lupus (GLADEL) cohort patients.⁴⁰

Quantification of the association between rates of flare and irreversible damage provides an opportunity to evaluate the impact of interventions upon this, offering potential value in health technology assessments. Several recent clinical trials have used flare as a secondary outcome measure. In the trials of belimumab, SLE Responder Index responders were shown to have reduced flares,⁴¹ and reduced flares were also observed in association with belimumab treatment when it was used after rituximab.⁴² Post hoc analysis of the phase 3 trials of anifrolumab reported significant reductions in flare.⁴³ In contrast, withdrawal of GC treatment in patients with stable inactive SLE was shown to be associated with increased flare,⁴⁴ as was detectable concentrations of serum interferon alpha.⁴⁵ Consistent with the alignment between overall reduction in disease activity and reduction in flare, no reduction in flare rate was observed in the negative Phase 3 trials of ustekinumab.⁴⁶

There are some limitations to this large patient cohort study. The APLC is not an ‘all-comers’ registry capturing all patients with SLE at a given location. It is possible that some patients with specific manifestations might have been managed in other specialty clinics, for example, nephrology. Differences in treatment approaches between countries and their corresponding health systems might limit the interpretation of data pooled across countries as well as limit the generalisability to other countries. Furthermore, while our study enumerates the association of damage outcomes with flare, it does not allow us to make specific recommendations regarding the use of pre-emptive treatments to reduce flares. Such recommendations can only come from interventional studies designed to address this question by measuring the effect of flare rate; it is our hope that the current data will aid in the design of such future studies.

The burden of damage in patients with SLE is substantial and flare has a quantifiable association with damage accrual. In SLE, damage accumulates over time and predicts further damage and mortality; therefore, early intervention to reduce the risk of damage is a critical therapeutic goal.²⁰ A multidimensional approach to damage prevention based on a better control of disease should include strategies to reduce flares and our findings that flares quantifiably worsen outcomes for patients with SLE suggest further evaluation of novel SLE therapies that reduce flares.

Data availability statement

Data are available upon reasonable request.