Frequency and Determinants of Flare and Persistently Active Disease in a Large Multinational Prospective Lupus Cohort

Abstract

Objective

In contrast to relapsing‐remitting patterns, persistently active disease (PAD) is a disease activity pattern in patients with systemic lupus erythematosus (SLE) that is inadequately studied. We sought to identify the frequency and determinants of flare and PAD in SLE.

Methods

Flare was defined using the Safety of Estrogens in Lupus Erythematosus National Assessment version of the Systemic Lupus Erythematosus Disease Activity Index (SELENA–SLEDAI flare index), and PAD was defined as an SLEDAI‐2K score of ≥4, excluding serology only, on two or more consecutive visits with a maximum six‐month interval. Multivariable logistic regression was used to develop predictive models for flare and PAD, which were tested in an independent validation subset.

Results

Among 3,811 patients over 2.8 (interquartile range 1.0–5.3) years of follow‐up, 2,142 (56.2%) experienced flare and 1,786 (46.9%) had PAD, with 368 (9.7%) experiencing PAD but not flare. The most common flare features were nephritis and arthritis, whereas PAD was most commonly characterized by renal or mucocutaneous activity. After adjusting for prednisone dose and use of antimalarials and immunosuppressants, low gross domestic product in country of residence, smoking, arthritis, nephritis, and low complement levels were predictive for flare, whereas being in a low disease activity state for ≥50% of follow‐up time (LLDAS50) was a protective factor. Renal activity and higher time‐adjusted mean SLEDAI‐2K were predictive of PAD, whereas LLDAS50 was protective. The models developed gave 72.1% and 83.8% correct classification of flare and PAD, respectively, in the validation cohort.

Conclusion

Both flare and PAD are common disease activity patterns in SLE; both predict organ damage accrual but differ in disease features and predictive factors. Because 9.7% of patients experience PAD but not flare, flare measures alone do not adequately capture all patients in whom disease control is suboptimal.

INTRODUCTION

Systemic lupus erythematosus (SLE) is an autoimmune condition characterized by variability in disease activity, with the relapsing-remitting pattern being the most commonly recognized pattern.^1–4 Recurrent exacerbations of SLE activity, usually referred to as flares and captured in cohort studies and clinical trials using flare-specific instruments, adversely impact long-term patient outcomes and lead to organ damage accrual.^5,6 Prevention of flares is therefore recommended as one of the key overarching principles and treatment goals in international recommendations for SLE management.^7,8 Although it has recently been highlighted that patients and clinicians conceive of flare differently, flare is commonly used as an outcome variable in SLE clinical trials.^9–13

In contrast to the relapsing-remitting pattern, some patients with SLE have periods of persistently active disease (PAD). However, unlike the relapsing-remitting pattern, PAD has been less commonly studied and is rarely used as an outcome in clinical trials. In a Canadian SLE cohort, Nikpour et al¹⁴ defined PAD as a Systemic Lupus Erythematosus Disease Activity Index 2000 (SLEDAI-2K) score of ≥4, excluding serological activity, on two or more consecutive visits, observing that periods of PAD were more common than flare episodes.

Although the frequency and determinants of flare in SLE have been reported in several studies,^5,15–17 the epidemiology of PAD is rarely reported and is not known in the Asia–Pacific region. The present study aims to identify and compare the frequency and determinants of flare and PAD and their association with damage accrual and to build and validate models that predict flare and PAD.

PATIENTS AND METHODS

Patients

This study included all patients in the Asia Pacific Lupus Collaboration (APLC) cohort enrolled between March 2013 and December 2020 who were aged ≥18 years old and had at least two visits during the follow-up period. The APLC cohort is a prospective, multinational, multicenter observational cohort of patients with SLE established in November 2012. Twenty-five sites across Australia, China, Japan, South Korea, Malaysia, the Philippines, Indonesia, Singapore, Taiwan, Thailand, New Zealand, and Sri Lanka have joined the APLC, with more than 4,000 patients recruited to date. Patients in the APLC cohort fulfill either the 1997 American College of Rheumatology modified classification criteria¹⁸ or the 2012 Systemic Lupus International Collaborating Clinics classification criteria for SLE.¹⁹ They can have newly diagnosed or pre-existing SLE, and their age must be 18 years and older. Patients are treated under standard care conditions and are followed up every three to six months. Standard-of-care treatments include antimalarials, glucocorticoids, nonsteroidal anti-inflammatory drugs, and immunosuppressants. Antimalarials are in principle prescribed for all patients unless contraindicated. Glucocorticoids and immunosuppressants, including biologics, are prescribed based on the patient's manifestations, disease activity, and severity. However, no specific treatment algorithm was predefined across the participating centers.²⁰

Ethics

Human research ethical approval for data collection, storage of the central data set, analysis, and publication of data collected by the APLC was obtained from the Monash University Human Research Ethics Committee (MUHREC Project ID 18778). Local ethics approvals have also been obtained at each participating site. Informed consent was obtained from each participant before enrollment.

Data collection

Patient demographics and SLE-related history were collected at enrollment. The SLEDAI-2K²¹; physician global assessment (PhGA; scale of 0–3)²²; medication use, including doses of glucocorticoids, antimalarial drugs, immunosuppressants (IS), and biologics; and disease flares were recorded at each visit. In addition, the Systemic Lupus International Collaborating Clinics/American College of Rheumatology Damage Index (SDI)²³ was completed annually. All data were recorded in a standardized electronic case report form.²⁰

Data availability statement

Access to APLC pooled data is subject to the specific guidelines outlined in the APLC Data Access Policy (available on request). The APLC welcomes requests for aggregate (summary) data or to perform analyses of new research questions, and such requests can be submitted to the APLC Steering Committee via the APLC Project Manager.

Definitions of flare, PAD, damage accrual, and lupus low disease activity state

Flare was defined using the Safety of Estrogens in Lupus Erythematosus National Assessment version of the Systemic Lupus Erythematosus Disease Activity Index (SELENA–SLEDAI flare index [SFI]).²⁴ Any visit that fulfilled the SFI definition was defined as a “flare visit” and was counted as one flare episode. PAD was defined as previously described, comprising a requirement of an SLEDAI-2K score of ≥4, excluding serological activity alone, on two or more consecutive visits with a maximum six-month interval between two visits.¹⁴ Within each episode of PAD, all consecutive visits, except for the first one, were defined as “PAD visits.”

Damage accrual was defined as any increase in the SDI²³ during follow-up. Lupus low disease activity state (LLDAS) was defined as: an SLEDAI-2K score of ≤4, with no activity in any major organ, no new disease activity since the previous assessment, a PhGA ≤1, a prednisolone (or equivalent) dose of ≤7.5 mg/day, and standard maintenance doses of IS and approved biologic agents.²⁵ LLDAS50 was defined as attaining LLDAS for ≥50% of follow-up time.

Statistical analysis

Continuous variable data are presented as medians with interquartile ranges (IQRs), and categorical variables are presented as percentages. The proportion of patients who had at least one episode of flare or PAD and the proportion of visits with flare or PAD were calculated. The incidence of flare or PAD is reported as the number of episodes of flare or PAD per patient-year. Organ-specific relapsing features at flare visits were reported according to the clinical domains in the definitions of the SFI, and organ-specific activities at PAD visits were reported according to the clinical domains in the definitions of the SLEDAI-2K.

Univariable and multivariable logistic regression analyses were used to analyze the association of flare and PAD with damage accrual over the follow-up period. Kaplan-Meier survival methods were used to evaluate cumulative damage accrual during follow-up, with the log-rank test applied for comparisons.

Predictive model development

We built prediction models for PAD and flare using data from patients who were enrolled in the APLC cohort from 2013 and had at least one visit each year from 2013 to 2016. Data from 2013 to 2015 were used to determine the odds ratio (OR) of flare or PAD in 2016. Age at diagnosis, sex, disease duration at enrollment, ethnicity, country of residence gross domestic product (GDP) per capita, education level, current smoking status, some disease activity variables during 2013 to 2015 (including organ domain-specific activity and serological activity according to the SLEDAI-2K, LLDAS50, time-adjusted mean SLEDAI-2K [AMS], and time-adjusted mean [TAM] PhGA), and treatment during 2013 to 2015 (including TAM prednisone daily dose, antimalarial use, cyclophosphamide [CYC] use, mycophenolate mofetil/mycophenolic acid [MMF/MPA] use, calcineurin inhibitor [CNI] use [including cyclosporine and tacrolimus], other IS use [including azathioprine, methotrexate, and leflunomide], rituximab [RTX] use, and belimumab [BEL] use) were included in the univariable logistic regression analysis. Based on the univariable analyses and clinical relevance, variables were included in the multivariable logistic regression to build the final model.

To test the predictive properties of the models, we used independent data from patients who were enrolled in the APLC cohort from 2017 and had at least one visit each year from 2017 to 2020. We used data from 2017 to 2019 to determine the ORs for flare or PAD in 2020 (Supplementary Figure 1). The predictive accuracy of each model analyzed is presented as sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), and overall percentage correct classification. It was also analyzed using receiver operating characteristic (ROC) curve analysis. All analyses were performed with STATA version 17.0 (StataCorp), and a P value of <0.05 was considered statistically significant.

RESULTS

Baseline characteristics of patients

There were 4,106 patients enrolled in the APLC cohort between March 2013 and December 2020, of whom 3,811 patients with 42,060 visits had more than one visit and were included in this study. Patients were predominantly female (92.1%) and Asian (89.1%). The median age was 29 (IQR 21–39) years at diagnosis and 39 (IQR 30–50) years at enrollment. The median disease duration at enrollment since SLE diagnosis was 8 (IQR 3–15) years, and the median follow-up duration was 2.8 (IQR 1.0–5.3) years. Renal activity, mucocutaneous involvement, and arthritis were the three most common clinical presentations, and positive anti–double-stranded DNA (anti-dsDNA) antibodies and low complement levels were found in 49.2% and 42.8% of the cohort, respectively, at recruitment. The median SLEDAI-2K score and PhGA at recruitment were 4 (IQR 2–6) and 0.5 (IQR 0.1–1), respectively (Table 1).

Table 1 Baseline and follow-up characteristics of all included patients*

Characteristics	Summary statistics (n = 3,811)
Baseline
Age at enrollment, median (IQR), y	39 (30–50)
Age at diagnosis, median (IQR), y	29 (21–39)
Female, n (%)	3,508 (92.1)
Disease duration at enrollment since diagnosis, median (IQR), y	8 (3–15)
Total visits per patient, median (IQR)	8 (5–15)
Race and ethnicity, n (%)	3,385 (88.8)
Asian	3,385/3,800 (89.1)
White	287/3,800 (7.5)
Other	128 (3.4)
Current smoker, n (%)	194/3,778 (5.1)
Tertiary education, n (%)	1,918/3,584 (53.5)
GDP (USD)
≥$50,000	1,906 (50.0)
$20,000–$50,000	559 (14.7)
<$20,000	1,346 (35.3)
Clinical features according to SLEDAI-2K, n (%)
Renal involvement	833 (21.9)
Mucocutaneous involvement	678 (17.8)
Arthritis	282 (7.4)
Leukopenia	185 (4.9)
Thrombocytopenia	106 (2.8)
NPSLE	44 (1.2)
Serositis	44 (1.2)
Fever	40 (1.1)
Vasculitis	40 (1.1)
Myositis	15 (0.4)
Serological profile according to SLEDAI-2K, n (%)
Positive anti-dsDNA antibodies	1,847/3,758 (49.2)
Low complement level	1,626/3,799 (42.8)
SLEDAI-2K score at enrollment, median (IQR)	4 (2–6)
PhGA score at enrollment, median (IQR)	0.5 (0.1–1)
Follow-up
Follow-up duration, median (IQR), y	2.8 (1.0–5.3)
Any flare ever, n (%)	2,142 (56.2)
Mild/moderate flare	2,031 (53.3)
Severe flare	987 (25.9)
Incidence of flare episode, per patient-year	0.56
PAD ever, n (%)	1,786 (46.9)
Incidence of PAD episode, per patient-year	0.30
LLDAS ever, n (%)	3,078/3,808 (80.8)
Percentage of time spent in LLDAS, median (IQR), %	50.0 (12.8–77.6)
TAM-PhGA, median (IQR)	0.4 (0.2–0.7)
AMS, median (IQR)	2.9 (1.3–4.7)
Medications
Prednisone use ever, n (%)	3,277 (86.0)
TAM prednisone dose, median (IQR), mg/day	5.0 (2.5–8.9)
Antimalarial use ever, n (%)	3,012 (79.0)
Conventional immunosuppressants use ever,a n (%)	2,725 (71.5)
Rituximab use ever, n (%)	82 (2.2)
Belimumab use ever, n (%)	68 (1.8)

Frequency of flare and PAD during follow-up

During follow-up, 2,142 (56.2%) and 1,786 (46.9%) patients experienced flare and PAD, respectively. There were 6,021 episodes (visits) of flare and 3,199 episodes of PAD covering 9,528 PAD visits. Among the 3,199 episodes of PAD, 1,496 episodes included only two consecutive visits, whereas 1,703 episodes included three or more consecutive visits. The incidences of flare and PAD were 0.56 and 0.30 episodes per patient-year, respectively (Table 1). Among the 3,811 patients, 1,301 (34.1%) patients experienced neither flare nor PAD, 724 (19.0%) patients had flare only, 368 (9.7%) patients experienced PAD only, and 1,418 (37.2%) patients experienced both flare and PAD (Supplementary Figure 2A).

On a visit level, neither flare nor PAD was present in 28,988 of 42,060 (68.9%) visits, whereas flare only was present in 3,544 (8.4%) visits, PAD only was present in 7,051 (16.8%) visits, and 2,477 (5.9%) visits fulfilled both flare and PAD definitions (Supplementary Figure 2B). The most common relapsing clinical features at the flare visits were nephritis (13.7% of flare visits) and arthritis (11.2%), whereas at the PAD visits, renal activity (71.2%) and mucocutaneous activity (23.6%) were the most common clinical manifestations (Table 2).

Table 2 Clinical features at visits with flare, PAD, or neither*

Organ involvement	Flare visitsa (n = 6,021)	PAD visitsb (n = 9,528)	Neither flare nor PAD visitsc (n = 28,988)
Renal, n (%)	833 (13.7)	6,785 (71.2)	1,433 (4.9)
Arthritis, n (%)	684 (11.2)	864 (9.1)	331 (1.1)
Mucocutaneous, n (%)	472 (7.8)	2,244 (23.6)	1,816 (6.2)
Leukopenia,d n (%)	-	681 (7.1)	733 (3.3)
Thrombocytopenia, n (%)	87 (1.4)	347 (3.6)	409 (1.8)
Fever, n (%)	88 (1.5)	40 (0.4)	23 (0.1)
Serositis, n (%)	88 (1.5)	133 (1.4)	66 (0.2)
Vasculitis, n (%)	84 (1.4)	142 (1.5)	43 (0.1)
Hemolytic anemia,e n (%)	75 (1.2)	–	–
CNS-SLE, n (%)	61 (1.0)	96 (1.0)	47 (0.2)
Myositis, n (%)	24 (0.4)	49 (0.5)	10 (0.03)

Association of flare and PAD with damage accrual

Univariable logistic regression analysis showed that both flare (OR 2.05, 95% confidence interval [CI] 1.71–2.46, P < 0.001) and PAD (OR 2.15, 95% CI 1.81–2.54, P < 0.001) were associated with damage accrual. In multivariable models, both flare (OR 1.29, 95% CI 1.04–1.60, P = 0.021) and PAD (OR 1.55, 95% CI 1.26–1.91, P < 0.001) maintained a statistically significant association with damage accrual after adjustment for age at diagnosis, disease duration and SLEDAI-2K score at enrollment, education level, ethnicity, follow-up duration, and TAM daily prednisolone dose (Table 3). Kaplan-Meier analysis also showed significantly higher cumulative damage accrual during follow-up in patients with flare compared to those without flare (log-rank P = 0.0001; Supplementary Figure 3A) and in patients with PAD compared to those without PAD (log-rank P = 0.0021; Supplementary Figure 3B).

Table 3 Association of flare and PAD with damage accrual during follow-up by logistic regression analysis*

	Univariable	Multivariablea
OR (95% CI)	P value	OR (95% CI)	P value
Never flare	Reference	Reference
Flare ever	2.05 (1.71–2.46)	<0.001	1.29 (1.04–1.60)	0.021
Never PAD	Reference	Reference
PAD ever	2.15 (1.81–2.54)	<0.001	1.55 (1.26–1.91)	<0.001

Prediction models for flare and PAD

We selected all patients enrolled in the APLC cohort from 2013 who had at least one visit each year from 2013 to 2016 (n = 803) to derive prediction models for flare and PAD. The baseline and follow-up characteristics of this group of patients (model derivation cohort) are shown in Supplementary Table 1. In univariable logistic regression analysis of data in the first three years (2013–2015), age at diagnosis, disease duration at enrollment, being a current smoker, GDP per capita of the country of residence, renal activity, arthritis, mucocutaneous involvement, serositis, leukopenia, low complement levels, positive anti-dsDNA antibodies, AMS score, LLDAS50, TAM prednisone daily dose, CYC use, MMF/MPA use, CNI use, other IS use, and RTX use had significant associations with flare in the fourth year (2016). However, BEL use was not associated with subsequent flare (Supplementary Table 2). Multivariable analysis showed that GDP <$20,000 US dollars (USD) (OR 2.80, 95% CI 1.65–4.76, P < 0.001), being a current smoker (OR 2.45, 95% CI 1.16–5.16, P = 0.019), renal activity (OR 3.53, 95% CI 2.29–5.44, P < 0.001), arthritis (OR 2.28, 95% CI 1.47–3.52, P < 0.001), low complement levels (OR 2.26, 95% CI 1.44–3.54, P < 0.001), and MMF/MPA use (OR 1.66, 95% CI 1.05–2.63, P = 0.029) in the first three years were risk factors for subsequent flare. LLDAS50 (OR 0.53, 95% CI 0.34–0.84, P = 0.007) was a protective factor against subsequent flare (Table 4).

Table 4 Prediction models for flare and PAD in 2016 derived from 2013–2015 data by multivariable logistic regression analysis*

Variables	Flarea	PADb
OR (95% CI)	P value	OR (95% CI)	P value
GDP <$20,000 USD	2.80 (1.65–4.76)	<0.001	–	–
Current smoker	2.45 (1.16–5.16)	0.019	–	–
Specific organ activity from 2013 to 2015 according to SLEDAI-2K
Renal	3.53 (2.29–5.44)	<0.001	3.76 (2.43–5.80)	<0.001
Arthritis	2.28 (1.47–3.52)	<0.001	–	–
Serological activity from 2013 to 2015 according to SLEDAI-2K
Low complement level	2.26 (1.44–3.54)	<0.001	–	–
AMS from 2013 to 2015	–	–	1.46 (1.29–1.65)	<0.001
Percentage of time spent in LLDAS50c from 2013 to 2015	0.53 (0.34–0.84)	0.007	0.45 (0.29–0.71)	<0.001
MMF/MPA use from 2013 to 2015	1.67 (1.05–2.63)	0.029	–	–

In univariable analysis for PAD, we found that the renal activity, arthritis, mucocutaneous involvement, leukopenia, low complement level, and positive anti-dsDNA antibodies SLEDAI-2K items; the AMS score; LLDAS50; TAM prednisone dose; CYC use; MMF/MPA use; CNI use; and other IS use in the first three years had significant associations with PAD in the fourth year. RTX and BEL use did not show an association with subsequent PAD (Supplementary Table 2). Multivariable regression showed that renal activity (OR 3.76, 95% CI 2.43–5.80, P < 0.001) and AMS score (OR 1.46, 95% CI 1.29–1.65, P < 0.001) were risk factors for PAD, whereas LLDAS50 (OR 0.45, 95% CI 0.29–0.71, P < 0.001) was a protective factor for PAD in year four (Table 4).

Properties of the prediction models for flare and PAD

We next tested these models in a validation cohort subset of patients enrolled in a separate period. All patients enrolled in the APLC cohort from 2017 who had at least one visit each year from 2017 to 2020 (n = 204) were used to test the prediction models for flare and PAD. The characteristics of this group of patients (model validation cohort) are shown in Supplementary Table 1. The median age of the patients in the validation cohort was slightly younger than that of those in the model derivation cohort (33 [IQR 26–44] vs 42 [IQR 33–52] years), the disease duration at enrollment was shorter (3 [IQR 1–7] vs 9 [IQR 5–15] years), and there were more patients with Asian ethnicity (94.6% vs 89.3%) and more patients from countries with a GDP <$20,000 USD (52.0% vs 24.3%).

When testing the model for prediction of flare, we found a sensitivity of 48.7% (95% CI 37.0%–60.4%), a specificity of 85.9% (95% CI 78.7%–91.4%), a PPV of 67.3% (95% CI 53.3%–79.3%), an NPV of 73.8% (95% CI 66.0%–80.7%), and an overall correct classification of 72.1% (Table 5). The area under the ROC curve for this model was 0.74 (Figure 1A).

Table 5 Properties of prediction models for flare and PAD in 2020 from the 2016–2019 data*

Model properties	Flare model	PAD model
Sensitivity (95% CI), %	48.7 (37.0–60.4)	51.0 (36.6–65.2)
Specificity (95% CI), %	85.9 (78.7–91.4)	94.8 (90.0–97.7)
Positive predictive value (95% CI), %	67.3 (53.3–79.3)	76.5 (58.8–89.3)
Negative predictive value (95% CI), %	73.8 (66.0–80.7)	85.3 (79.1–90.3)
Overall correct prediction, %	72.1	83.8

[IMAGE OMITTED. SEE PDF]

When testing the model for the prediction of PAD, we found a sensitivity of 51.0% (95% CI 36.6%–65.2%), a specificity of 94.8% (95% CI 90.0%–97.7%), a PPV of 76.5% (95% CI 58.8%–89.3%), an NPV of 85.3% (95% CI 79.1%–90.3%), and an overall correct classification of 83.8% (Table 5). The area under the ROC curve for this model was 0.88 (Figure 1B).

DISCUSSION

This study evaluated the frequency and predictors of PAD and flare in a large multinational cohort of patients with SLE. In this cohort, more than half of the patients (56.2%) experienced at least one flare during a median follow-up of 2.8 years, with an incidence of 0.56 per patient-year, comparable with some previous reports.^{16,17,26–30} PAD was also commonly observed, with nearly half of all patients (46.9%) in our cohort experiencing PAD. In fact, we found that 37.2% of the patients experienced episodes of both flare and PAD during follow-up and 9.7% of the patients only had PAD without experiencing flare during their follow-up. In the study from the Toronto Lupus Cohort, Nikpour et al¹⁴ reported that at least 25% of patients had PAD without achieving the definition of flare. Both the Toronto study and ours confirm that PAD is a common disease activity pattern in SLE.

This study also reported the association of flare and PAD with damage accrual, showing that both flare and PAD are significantly associated with damage accrual. The high frequency of PAD and its negative impact on patient outcomes make a compelling case for its use as a clinically meaningful end point.

We used independent subcohorts of patients enrolled in different time periods to develop and test the properties of prediction models for flare and PAD. Although we used an internal validation cohort, there was no overlap between these two subcohorts. Indeed, there were some differences in patients’ demographic characteristics, such as age, disease duration, ethnicity, and country of origin, suggesting that the developed models might work across different cohorts.

We noted that the renal activity had the strongest association with flare and also PAD. Several studies have also reported the predictive association of renal involvement and flare.^15,17 However, this study also reported the impact of renal activity on subsequent PAD, further highlighting the poorer prognosis of patients with lupus nephritis and the importance of care and follow-up of these patients. We also found that hypocomplementemia was a risk factor for flare (OR 2.19, 95% CI 1.46–3.30, P < 0.001) but not for PAD. Some previous reports have shown the association of serological activity, such as positive anti-dsDNA antibodies and low complement levels, with an increased risk of flare.^15,26,31 Our result showed that low complement levels have a stronger association with flare than anti-dsDNA. Regarding treatment, higher TAM prednisone dose, different IS use, and RTX use were associated with a higher chance of flare in univariable analysis, and MMF/MPA use was associated with more flares in the multivariable flare model. These associations likely indicate that the requirement for more intensive treatment is a signal for having a higher chance of future flare. In the PAD model, another identified predictive factor was prior TAM disease activity. Higher prior disease activity predicts a likelihood of subsequently experiencing PAD, possibly pointing to an inherently more difficult-to-treat subgroup or suboptimal treatment adherence. In addition, both of our flare and PAD models showed that LLDAS50 in the preceding three years was associated with a reduced risk of developing flare or PAD in a fourth year, further confirming the benefit of maintaining a low disease activity status for as long as possible; the protective effects of sustained LLDAS have recently been formally confirmed.³² Our models performed well during validation, especially for PAD. These models could assist physicians in identifying patients at high risk of flare and PAD who require stringent attention to controlling disease activity to avoid adverse outcomes. They could also be used to inform patient selection for clinical trials.

There are limitations of this study. Firstly, the definition of PAD was based on the SLEDAI-2K, which has some inherent limitations. For example, the SLEDAI-2K omits several clinical manifestations, including hemolytic anemia and some cardiac, pulmonary, and gastrointestinal manifestations. Therefore, patients with these clinical features are not captured by an SLEDAI-2K-based definition of PAD. On the other hand, the SLEDAI-2K-based definition may overestimate the frequency of PAD. Several items in the SLEDAI-2K definition need not be “new” or “recurrent,” including rash, alopecia, mucosal ulcers, and proteinuria. Persistent manifestations in these items can also be scored. Some of these persistent manifestations may not necessarily represent real active disease. For example, some studies showed that proteinuria can be discordant and lag behind histologic remission in lupus nephritis.³³ Therefore, renal activity may be overestimated in this case by using the SLEDAI-2K. Secondly, association does not prove causation, and future prospective intervention studies that impact risk factors for PAD and flare should be used to assess the performance of our models. Thirdly, there was very limited use of RTX and BEL in this cohort, and no data were available on the use of anifrolumab and ciclosporin. An updated analysis with more frequent use of these relatively newer medications would be interesting to conduct in the future. Notwithstanding this, performing this study in a very large prospective multinational longitudinal cohort and using derivation and validation cohort subsets to develop and validate our models are strengths.

In conclusion, we have shown that the frequency of flare and PAD is substantial under standard-of-care conditions and that some patients only experience PAD despite not having flare, at least as defined using the SFI. Importantly, we have shown that PAD, like flare, is associated with an increased risk of damage accrual. We have also shown that active disease in the renal domain is a risk factor for developing both flare and PAD but spending more than 50% of the time in LLDAS is a protective factor against future flare and PAD, highlighting the importance of stringent disease control in improving outcomes. Our predictive models may help identify patients at high risk of flare or PAD to target for more stringent follow-up and assist in selecting high-risk patients for clinical trials.

ACKNOWLEDGMENTS

We thank all the patients enrolled in the APLC for their participation. Open access publishing facilitated by The University of Melbourne, as part of the Wiley - The University of Melbourne agreement via the Council of Australian University Librarians.

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 Nikpour 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.

References

Györi N, Giannakou I, Chatzidionysiou K, et al. Disease activity patterns over time in patients with SLE: analysis of the Hopkins Lupus Cohort. Lupus Sci Med 2017;4(1): [eLocator: e000192].

Zen M, Bassi N, Nalotto L, et al. Disease activity patterns in a monocentric cohort of SLE patients: a seven‐year follow‐up study. Clin Exp Rheumatol 2012;30(6):856–863.

Barr SG, Zonana‐Nacach A, Magder LS, et al. Patterns of disease activity in systemic lupus erythematosus. Arthritis Rheum 1999;42(12):2682–2688.

Tselios K, Gladman DD, Touma Z, et al. Disease course patterns in systemic lupus erythematosus. Lupus 2019;28(1):114–122.

Ugarte‐Gil MF, Acevedo‐Vásquez E, Alarcón GS, et al; GLADEL. The number of flares patients experience impacts on damage accrual in systemic lupus erythematosus: data from a multiethnic Latin American cohort. Ann Rheum Dis 2015;74(6):1019–1023.

Koutsonikoli A, Trachana M, Heidich AB, et al. Dissecting the damage in Northern Greek patients with childhood‐onset systemic lupus erythematosus: a retrospective cohort study. Rheumatol Int 2015;35(7):1225–1232.

van Vollenhoven RF, Mosca M, Bertsias G, et al. Treat‐to‐target in systemic lupus erythematosus: recommendations from an international task force. Ann Rheum Dis 2014;73(6):958–967.

Fanouriakis A, Kostopoulou M, Andersen J, et al. EULAR recommendations for the management of systemic lupus erythematosus: 2023 update. Ann Rheum Dis 2024;83(1):15–29.

Morand EF, Jones SA. Flare in systemic lupus erythematosus: lost in translation? J Rheumatol 2024;51(5):437–439.

Furie R, Petri M, Zamani O, et al; BLISS‐76 Study Group. A phase III, randomized, placebo‐controlled study of belimumab, a monoclonal antibody that inhibits B lymphocyte stimulator, in patients with systemic lupus erythematosus. Arthritis Rheum 2011;63(12):3918–3930.

Navarra SV, Guzmán RM, Gallacher AE, et al; BLISS‐52 Study Group. Efficacy and safety of belimumab in patients with active systemic lupus erythematosus: a randomised, placebo‐controlled, phase 3 trial. Lancet 2011;377(9767):721–731.

Morand EF, Furie R, Tanaka Y, et al; TULIP‐2 Trial Investigators. Trial of anifrolumab in active systemic lupus erythematosus. N Engl J Med 2020;382(3):211–221.

Furie R, Petri MA, Strand V, et al; BLISS‐52 and BLISS‐76 Study Groups. Clinical, laboratory and health‐related quality of life correlates of Systemic Lupus Erythematosus Responder Index response: a post hoc analysis of the phase 3 belimumab trials. Lupus Sci Med 2014;1(1): [eLocator: e000031].

Nikpour M, Urowitz MB, Ibañez D, et al. Frequency and determinants of flare and persistently active disease in systemic lupus erythematosus. Arthritis Rheum 2009;61(9):1152–1158.

Petri MA, van Vollenhoven RF, Buyon J, et al; BLISS‐52 and BLISS‐76 Study Groups. Baseline predictors of systemic lupus erythematosus flares: data from the combined placebo groups in the phase III belimumab trials. Arthritis Rheum 2013;65(8):2143–2153.

Petri M, Singh S, Tesfasyone H, et al. Prevalence of flare and influence of demographic and serologic factors on flare risk in systemic lupus erythematosus: a prospective study. J Rheumatol 2009;36(11):2476–2480.

Inês L, Duarte C, Silva RS, et al. Identification of clinical predictors of flare in systemic lupus erythematosus patients: a 24‐month prospective cohort study. Rheumatology (Oxford) 2014;53(1):85–89.

Hochberg MC. Updating the American College of Rheumatology revised criteria for the classification of systemic lupus erythematosus. Arthritis Rheum 1997;40(9):1725.

Petri M, Orbai AM, Alarcón GS, et al. Derivation and validation of the Systemic Lupus International Collaborating Clinics classification criteria for systemic lupus erythematosus. Arthritis Rheum 2012;64(8):2677–2686.

Kandane‐Rathnayake R, Golder V, Louthrenoo W, et al. Development of the Asia Pacific Lupus Collaboration cohort. Int J Rheum Dis 2019;22(3):425–433.

Ibañez D, Gladman DD, Urowitz MB. Adjusted mean Systemic Lupus Erythematosus Disease Activity Index‐2K is a predictor of outcome in SLE. J Rheumatol 2005;32(5):824–827.

Petri M, Hellmann D, Hochberg M. Validity and reliability of lupus activity measures in the routine clinic setting. J Rheumatol 1992;19(1):53–59.

Gladman DD, Goldsmith CH, Urowitz MB, et al. The Systemic Lupus International Collaborating Clinics/American College of Rheumatology (SLICC/ACR) damage index for systemic lupus erythematosus international comparison. J Rheumatol 2000;27(2):373–376.

Petri M, Buyon J, Kim M. Classification and definition of major flares in SLE clinical trials. Lupus 1999;8(8):685–691.

Golder V, Kandane‐Rathnayake R, Huq M, et al; Asia‐Pacific Lupus Collaboration. Lupus low disease activity state as a treatment endpoint for systemic lupus erythematosus: a prospective validation study. Lancet Rheumatol 2019;1(2):e95–e102.

Floris A, Piga M, Cauli A, et al. Predictors of flares in systemic lupus erythematosus: preventive therapeutic intervention based on serial anti‐dsDNA antibodies assessment. Analysis of a monocentric cohort and literature review. Autoimmun Rev 2016;15(7):656–663.

Petri M, Genovese M, Engle E, et al. Definition, incidence, and clinical description of flare in systemic lupus erythematosus. A prospective cohort study. Arthritis Rheum 1991;34(8):937–944.

van den Berg L, Nossent H, Rekvig O. Prior anti‐dsDNA antibody status does not predict later disease manifestations in systemic lupus erythematosus. Clin Rheumatol 2006;25(3):347–352.

Cervera R, Doria A, Amoura Z, et al. Patterns of systemic lupus erythematosus expression in Europe. Autoimmun Rev 2014;13(6):621–629.

Conti F, Ceccarelli F, Perricone C, et al. Flare, persistently active disease, and serologically active clinically quiescent disease in systemic lupus erythematosus: a 2‐year follow‐up study. PLoS One 2012;7(9): [eLocator: e45934].

Hao Y, Ji L, Gao D, et al. Flare rates and factors determining flare occurrence in patients with systemic lupus erythematosus who achieved low disease activity or remission: results from a prospective cohort study. Lupus Sci Med 2022;9(1): [eLocator: e000553].

Golder V, Kandane‐Rathnayake R, Li N, et al. Association of sustained lupus low disease activity state with improved outcomes in systemic lupus erythematosus: a multinational prospective cohort study. Lancet Rheumatol 2024;6(8):e528–e536.

Malvar A, Pirruccio P, Alberton V, et al. Histologic versus clinical remission in proliferative lupus nephritis. Nephrol Dial Transplant 2017;32(8):1338–1344.