No benefit of adding glucocorticoids to maintenance treatment in reducing the risk of major relapses in ANCA-associated vasculitis: real life data from a longitudinal cohort study

Abstract

Objective

To examine the role of low-dose glucocorticoids (GCs) in major relapse, hospitalisation and damage accumulation risk during maintenance therapy in anti-neutrophil cytoplasmic antibody (ANCA)-associated vasculitis (AAVs).

Methods

Retrospective cohort study of newly diagnosed patients with granulomatosis with polyangiitis (GPA) or microscopic polyangiitis (MPA) followed in three tertiary referral centres. We recorded relapses (Birmingham Vasculitis Activity Score increase >0), increases in Vasculitis Damage Index (VDI) and hospitalisations. We used time-varying and mixed-effects Cox models to examine the effect of GCs on the risk of major relapses, VDI accumulation and hospitalisations. Sensitivity analysis was also conducted to address potential confounding.

Results

171 patients (GPA: 107, MPA: 64, median age: 61 years) were included with a median follow-up of 88.2 months (803.4 patient years (PY)). We recorded 65 major relapses (8.1/100 PY) in 48 patients (28%), 65 events of new damage (8.1/100 PY) and 132 hospitalisations (16.4/100 PY). By multivariable analysis, rituximab use was associated with a lower risk (HR=0.21, 95% CI 0.09 to 0.47, p=0.0001) while disease activity at diagnosis was associated with an increased risk (HR=1.17, 95% CI 1.03 to 1.33, p=0.013) of major relapses. Low-dose GCs were not associated with a reduced major relapse risk (HR=0.96, 95% CI 0.88 to 1.05, p=0.36) but they were associated with a risk of damage accumulation (HR=1.52, 95% CI 1.19 to 1.93, p=0.0007) and hospitalisations (HR=1.24, 95% CI 1.06 to 1.45, p=0.006).

Conclusion

Higher GC exposure during maintenance therapy was not significantly associated with a lower major relapse risk whereas it was associated with damage accrual and hospitalisation. These findings may support decision making regarding long-term GC use in patients with GPA/MPA.

WHAT IS ALREADY KNOWN ON THIS TOPIC

• Rituximab during maintenance treatment in patients with anti-neutrophil cytoplasmic antibody-associated vasculitis reduces the risk of relapses.

• There are conflicting data regarding the value of low-dose glucocorticoids (GCs) during maintenance therapy in preventing major relapses.

WHAT THIS STUDY ADDS

• This long-term, real life study provides evidence that low-dose GCs added to standard maintenance agents (rituximab and non-rituximab ones) were not significantly associated with a lower risk for major relapses while they were associated with damage accrual and hospitalisations.

HOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE OR POLICY

• In patients with granulomatosis with polyangiitis/microscopic polyangiitis who achieve clinical remission, attempts should be made to gradually taper and discontinue GCs, avoiding long-term patient exposure and adverse events.

Introduction

Anti-neutrophil cytoplasmic antibody (ANCA)-associated vasculitides (AAV) are rare diseases with considerable damage, morbidity and mortality due to the disease itself, comorbidities and/or its treatment.^{1 2} The main goals of treatment include achievement of initial remission and prevention of relapses and chronic damage during the maintenance of remission period.^{3 4} The use of cyclophosphamide (CYC) or rituximab (RTX) in combination with glucocorticoids (GCs) as induction regimens has significantly reduced mortality at approximately 10%–25% (from 93% during the first 2 years in the pre-CYC era), with 80% of patients achieving remission during the first year.^{5 6}

Regarding maintenance therapy, randomised controlled trials (RCTs)⁷,⁹ and real life studies¹⁰,¹² have shown that RTX was superior to azathioprine (AZA) in preventing relapses. Our recent work has shown that relapses occurred in 24% of patients receiving RTX maintenance therapy with an incidence rate (IR) of 10.2/100 patient years (PY) with most relapses (73%) occurring the first 2 years.¹³ Recently, both the European Alliance of Associations for Rheumatology (EULAR) and the American College of Rheumatology (ACR) proposed RTX as the preferred agent for maintenance of remission in patients with AAV such as granulomatosis with polyangiitis (GPA) and microscopic polyangiitis (MPA).^{3 4}

While the role of RTX in reducing relapses is clearly established, data on the effect of GCs during the maintenance period are contradictory. An older meta-analysis has shown that patients who continued GCs had less relapses (14%) compared with those who had discontinued them (43%),¹⁴ while in contrast, a retrospective study of 147 patients demonstrated that patients who had received GCs for more than 6 months had the same relapse risk as those who had stopped them.¹⁵

In view of these limited data, the aim of our study was to examine the effect of low-dose GCs on relapse risk (especially major relapses) once remission with an induction regimen in newly diagnosed patients with GPA or MPA had been achieved. Since long-term GC use can be associated with adverse effects such as damage accumulation or hospitalisations, the effect of GCs on their risk was also investigated.

Methods

Patient population, data collection and study design

We conducted a retrospective cohort study with longitudinally collected data from three major referral centres. We included newly diagnosed patients with GPA or MPA, according to Chapel Hill Consensus Conference definitions,¹⁶ who had achieved remission (as described in ‘Definitions’) at least once after induction therapy and had at least two subsequent consecutive visits. Patients with eosinophilic granulomatosis with polyangiitis (EGPA) or with missing data concerning treatment were excluded (online supplemental figure 1).

Data were obtained with visit-by-visit evaluations, including all the routine visits, urgent visits and dates of hospital admissions. GC dose, disease activity (assessed by the Birmingham Vasculitis Activity Score (BVAS)), disease damage (assessed by the Vasculitis Damage Index (VDI)) and hospitalisations were recorded at every visit. For the recording of GC use, all routes of administration (oral, intravenous or intramuscular) were taken into account, and all different formulations were converted to equivalent doses of prednisolone. Subsequently, the mean daily GC (prednisolone) dose was calculated every time for the period between consecutive visits.

Concerning baseline and demographic characteristics, age, sex, induction treatment, comorbidities at the diagnosis (using Charlson Comorbidity Index), BVAS at the diagnosis, VDI at the time of remission onset, manifestations of the disease and ANCA serotype were also recorded. Sex-related data were physician-reported.

Definitions

We used the following definitions:

Remission: A period during which the BVAS is equal to zero, the patient receives prednisolone equivalents <7.5 mg/day and has completed CYC or RTX induction treatment.

Major relapse: Any increase in BVAS from zero with a major organ involvement or a life-threatening manifestation, treated with major immunosuppression or high dose of GCs. Treatment escalation provided additional supportive evidence of the clinical severity of the event.

Minor relapse: Any increase in BVAS from zero that does not meet the criteria for a major relapse.

Kidney involvement: Histopathological findings in the kidney biopsy and/or a raise in serum creatinine and/or red blood cell casts and/or significant glomerular haematuria in the urine sediment and/or proteinuria >500 mg.

Refractory disease: Unchanged/increased signs and/or symptoms of active disease after a period of induction therapy, as long as other potential causes of persistent/worsened disease manifestations were ruled out.

Hospitalisations: Clinically relevant admissions during maintenance follow-up, mainly related to infections, disease activity, treatment-related complications, cardiovascular or renal complications, and other comorbid events occurring in the context of AAV management. Clearly unrelated or incidental admissions were not systematically considered as hospitalisation outcomes in the present analysis.

Statistical analysis

Statistical analysis was based on survival analysis and was conducted in two parts. The first part referred to the possible association between time-updated GCs (mean daily dose between consecutive visits) during maintenance follow-up and the hazard of (first) major relapse, (first) VDI increase or (first) hospitalisation. We also explored a composite outcome of VDI accumulation and hospitalisations, using a proportional hazards Cox model with time-varying variables. The second part referred to the effect of GCs on major flares, major and minor flares overall and hospitalisations, considering all the outcomes (in patients with multiple events), by the use of a mixed-effects Cox model for multiple events, employing a gap-time approach for modelling the waiting times. Importantly, no predefined treatment strategy comparing continuous versus intermittent GC administration was evaluated in this study.

The multivariable models were created based on a combination of clinical relevance, previously reported predictors of relapse and adverse outcomes in AAV and the primary objectives of the study. Candidate variables included demographic characteristics, BVAS at diagnosis, induction treatments, disease and ANCA phenotype, cumulative GC dose during induction, as well as maintenance treatments (RTX vs non-RTX agents vs no therapy), BVAS at every visit, VDI at every visit and hospitalisations at every visit (time-varying variables). Furthermore, a parsimonious modelling approach was adopted, taking into account the number of events relative to the number of candidate variables. Variables whose inclusion did not alter the estimates or statistical significance of the remaining covariates were not retained in the final models, in order to reduce the risk of overparameterisation and model instability. The results of these models were presented as HRs. The proportionality of hazards assumption was checked using the Shoenfeld residuals for every model. For the modelling of multiple VDI increases, a Generalized Linear Mixed Poisson Model (GLMM) was employed, as VDI can increase >1 point at each time. For this purpose, the difference of each VDI Score from the previous one was calculated. The results of this model were presented as rate ratios. The logarithm of time, from one visit to the next one, was introduced in the model as an offset.

Time-to-event outcomes concerning the first major relapse, VDI increase and hospitalisation were analysed using the Kaplan-Meier estimator to calculate and visualise unadjusted survival probabilities over time. Also, adjusted survival curves were derived from the Cox models to visualise covariate-adjusted survival probabilities concerning the first major relapse. Major relapse-free survival curves were stratified by GC use and RTX maintenance therapy and were generated with the covariates of the model fixed at representative values (median age and median BVAS at diagnosis were used as well as median GC daily dose for the survival curves stratified for RTX maintenance therapy). To address potential time-dependent confounding affected by the possible relationship between evolving disease activity (BVAS), prior treatment exposure and subsequent treatment decisions, we performed a sensitivity analysis using a marginal structural Cox model with inverse probability of treatment weighting (MSM/IPTW). At each visit interval, a logistic regression model was used to estimate the probability of receiving GC, conditional on prior GC exposure, lagged BVAS and VDI, prior hospitalisations, maintenance treatment, as well as baseline characteristics (age, cumulative GC at induction, BVAS at diagnosis). Individual weights were multiplied across follow-up intervals to obtain cumulative weights. Stabilised weights were truncated at the 5th and 95th percentiles to reduce the influence of extreme observations. Finally, the weighted Cox model was used to estimate the association between GC exposure and outcomes (major relapses, VDI accumulation and hospitalisations).

For descriptive statistics, median (IQR) and absolute and relative frequencies were used for continuous variables and categorical variables, respectively. Finally, statistical significance was considered for values of p<0.05. Analyses were performed using R Statistical Software (V.4.3.3; R Core Team 2024).

Results

Patient characteristics: a longitudinal cohort with ~7 years of median follow-up

Overall, 171 patients with GPA (n=101) and MPA (n=64) were included (table 1). The patients were followed for a median time of 88.2 (IQR: 74.2) months with 2295 visits and 803.4 PY of follow-up. The patient characteristics are shown in table 1. Their median age at diagnosis was 61 years while 88 (51.5%) patients were female. The majority were ANCA+ (89%, 47% C-/Proteinase-3 ANCA+, 42% P-/Myeloperoxidase ANCA+). Most patients had lung (n=119, 69.6%) and kidney (n=73, 64.3%) involvement at diagnosis; among those with kidney involvement 57 (78%) had a kidney biopsy. Of note, 119 (69.6%) patients had more than one comorbidity at diagnosis, while almost half of them (n=83, 48.5%) had accumulated more than one non-reversible manifestation (damage) by the end of the induction period (table 1). The full spectrum of clinical manifestations during their disease course is presented in online supplemental table 1.

Table 1. Baseline patient and treatment characteristics.

Characteristic
Age at diagnosis (years)	61 (23.9)
Sex
Female	88 (51.5)
AAV type
GPA	107 (62.6)
MPA	64 (37.4)
ANCA status
cANCA/PR3+	80 (47)
pANCA/MPO+	72 (42)
ANCA−	19 (11)
BVAS at diagnosis	14 (9)
Organ/system involvement at diagnosis
Lungs	119 (69.6)
General	116 (67.8)
Kidneys	110 (64.3)
Ear, nose and throat	73 (42.7)
Peripheral nerves	28 (16.4)
Skin	19 (11.1)
Eyes	10 (5.8)
Central nervous system	9 (5.3)
Heart	2 (1.2)
Gastrointestinal	1 (0.06)
Charlson Comorbidity Index at diagnosis
0	52 (30.4)
1	20 (11.7)
2	26 (15.2)
3	23 (13.5)
4	26 (15.2)
≥5	24 (14)
Induction treatment
CYC only	54 (31.6)
RTX only	68 (39.8)
CYC+RTX*	15 (9.7)
CYC/RTX switched to CYC/RTX†	17 (9.9)
AZA	7 (4.1)
MTX	16 (9.4)
Avacopan	6 (3.5)
MMF	1 (0.6)
Cumulative GC dose during induction (gm)	5.8 (3.3)
VDI at the end of induction
0	72 (42.1)
1	59 (34.5)
2	22 (12.9)
≥3	18 (10.5)
Time to first remission (days)	194 (127.5)

Data are n (%) or median (IQR).

^*At the same time.

^†Consecutively due to non-response (refractory cases).

.AAV, ANCA-associated vasculitis; ANCA, Anti-neutrophil cytoplasmic antibody; AZA, azathioprine; BVAS, Birmingham Vasculitis Activity Score; cANCA, cytoplasmic ANCA; CYC, cyclophosphamide; GC, glucocorticoid; GPA, granulomatosis with polyangiitis; MMF, mycophenolate mofetil; MPA, microscopic polyangiitis; MPO, myeloperoxidase; MTX, methotrexate; pANCA, perinuclear ANCA; PR3, proteinase 3; RTX, rituximab; VDI, Vasculitis Damage Index.

Induction and maintenance treatment characteristics: comparable exposure of patients to RTX versus non-RTX agents and to GC versus GC-free maintenance

With regard to induction schemes, 90.1% (n=154) of patients received RTX, CYC or both (overall CYC: n=86, 50.3%, RTX: n=100, 58.5%) in combination with GCs at a median (IQR) cumulative dose of 5.8 (3.3) gm (table 1).

After the first remission was achieved, 153 (89.5%) patients received RTX at least once, while the median (IQR) time of RTX treatment overall was 24.3 (40.8) months. Non-RTX agents were used as maintenance in 69 (40.3%) patients with a median (IQR) time of exposure of 28.8 (42) months. Finally, 71 (41.5%) patients stopped at least once the maintenance treatment and the median (IQR) off treatment time was 8.4 (26.8) months.

With regard to GC use, 152 (88.9%) patients received them at least once during maintenance. Among them, 44 patients had GC exposure before their first major flare while 4 additional patients received them after their first major flare (figure 1). The median (IQR) time of overall exposure was 17.5 (28.7) months, while the median (IQR) time that patients were off GCs was 28.7 (58.5) months, yielding comparable treatment groups. Among patients who received GCs, 129/152 (84.9%) patients were able to discontinue GCs at least once (figure 1).

Figure 1. Swimmer plot showing the first and subsequent major relapses in individual patients who were using: (A) Continuously (n=22), (B) Intermittently (n=22) or (C) No (n=4), GCs until their first major relapse*. *Light blue segments indicate periods on GCs treatment and dark blue segments indicate periods off GCs. Red crosses represent major relapses. ANCA, anti-neutrophil cytoplasmic antibody; AAV, ANCA-associated vasculitis; GC, glucocorticoids.

Outcomes

Disease relapses: GC use during maintenance does not provide an additional benefit in preventing major relapses

During follow-up, 83 patients (48.5%) experienced at least one relapse; among them 48 (28%) had at least one major and 57 (32%) a minor relapse. Overall, 65 major (IR: 8.1/100 PY) and 86 minor (IR: 10.7/100 PY) relapses were recorded. In figure 2A, the Kaplan-Meier curve for the first major relapse-free survival for the whole patient cohort is shown.

Figure 2. (A) Kaplan-Meier estimate of time to first major relapse for the whole patient cohort. (B,C) Predicted first major relapse-free survival curves according to RTX (B), and GC (C) use, during the maintenance period, respectively. Censored data are noted with a cross on the Kaplan-Meier curves. Both curves (Β, C) were based on the fitted Cox model. Survival curves were generated with the covariates of the model fixed at representative values for median age and median BVAS at diagnosis (B) and median age, median GC daily dose and median BVAS at diagnosis (C), respectively. ANCA, anti-neutrophil cytoplasmic antibody; AAV, ANCA-associated vasculitis; BVAS, Birmingham Vasculitis Activity Score; GC, glucocorticoids; RTX, rituximab.

The characteristics of patients with or without relapses are shown in online supplemental table 2. In general, patients who relapsed had more often GPA and less often received RTX as induction therapy compared with non-relapsers. Most major relapses (45/65, 69.2%) occurred in the same organs as the initial involvement. New organs/systems affected more often were the ear, nose and throat (ENT, 5/15, 33.3%), kidneys (4/15, 26.7%), nervous system (4/15, 26.7%) and lungs (3/15, 20%). The full spectrum of manifestations, both at disease onset and at relapses, is presented in online supplemental table 3.

The pattern of GC exposure of patients who developed (n=48) or not (n=123) at least one major relapse is shown in figure 1 and online supplemental figure 2, respectively. Overall, 53 patients received GCs continuously (31%), 95 (56%) intermittently, while 23 (13%) were never exposed to GCs after their first remission. At the time of the first major relapse, a similar number of patients were on (48%, n=23) or off (52%, n=25) GCs (figure 1).

Regarding the risk for the first major relapse by multivariable analysis, RTX use as maintenance was associated with a lower risk (HR=0.21, 95% CI 0.09 to 0.47, p=0.0001, table 2, figure 2B) while increased disease activity at diagnosis (BVAS at diagnosis, for every three units increase: HR 1.17, 95% CI 1.03 to 1.33, p=0.013, table 2) was associated with an increased risk.

Table 2. Multivariable model for the risk of the first major relapse, VDI accumulation, hospitalisation and composite outcome (hospitalisation with VDI accumulation) during the maintenance period.

Outcome	Covariates	HR	P value	95% CI
Major relapses	GC maintenance (mg/day)	0.96	0.36	0.88 to 1.05
	RTX maintenance	0.21	0.0001	0.09 to 0.47
	Non-RTX maintenance	0.71	0.36	0.34 to 1.48
	Age (years)	0.99	0.77	0.98 to 1.02
	BVAS at diagnosis	1.17*	0.013	1.03 to 1.33
VDI accumulation	GC maintenance (mg/day)	1.52†	0.0007	1.19 to 1.93
	RTX maintenance	1.07	0.82	0.57 to 2.04
	BVAS at every visit	1.42	<0.0001	1.25 to 1.61
	Interaction term BVAS x GCs maintenance	0.89	0.0003	0.84 to 0.95
	Age (years)	1.10‡	0.0077	1.03 to 1.18
	Hospitalisations	1.79	0.43	0.42 to 7.62
Hospitalisations	GC maintenance (mg/day)	1.24†	0.0060	1.06 to 1.45
	RTX Maintenance	1.23	0.49	0.68 to 2.22
	BVAS at every visit	1.14	0.10	0.99 to 1.3
	Interaction term BVAS x GCs maintenance	0.98	0.058	0.95 to 0.99
	Age (years)	1.02	0.10	0.99 to 1.04
	Non-RTX maintenance	1.04	0.91	0.50 to 2.20
	VDI at every visit	1.32	0.0022	1.10 to 1.57
Composite outcome	GC maintenance (mg/day)	1.11†	0.0008	1.04 to 1.18
	RTX maintenance	1.35	0.27	0.76 to 2.40
	BVAS at every visit	1.13	0.0039	1.04 to 1.23
	BVAS at diagnosis	1.14*	0.017	1.02 to 1.27
	Age (years)	1.22‡	0.0068	1.03 to 1.19
	Non-RTX maintenance	1.23	0.54	0.63 to 2.41

Statistically significant variables are displayed in bold. Adjustment for the different hospitals and cumulative dose of GCs during the induction period of treatment has been included in the analysis.

^*For every 3 units increase.

^†For every 5 mg increase.

^‡For every 5 years increase.

BVAS, Birmingham Vasculitis Activity Score; GC, glucocorticoids; RTX, rituximab; VDI, Vasculitis Damage Index.

GC use during the maintenance period did not reduce the risk for relapses (HR=0.96, 95% CI 0.88 to 1.05, p=0.36, table 2 and figure 2C).

In order to assess the risk for patients who had more than one major relapse, a mixed-effects model was used. By this analysis, GC use was not associated with a reduced risk either for major (HR=0.96, p=0.43) or all (major and minor) relapses (HR=0.95, p=0.12, online supplemental table 4), while RTX use during maintenance was associated with a lower (HR=0.3, p=0.0049) risk. Similarly to the multivariable analysis for the first major relapse, high disease activity at diagnosis (BVAS) was associated with a higher risk for major relapses (for every three units increase: HR=1.82, p=0.0062, online supplemental table 4).

Disease damage: chronic GC use increases the risk of damage accumulation

At the remission onset (beginning of follow-up), damage accumulation was already observed, with 88 components occurring due to vasculitis-related damage, 18 due to GC toxicity and 5 due to comorbidities (online supplemental table 5). During the maintenance period, an increase in VDI was observed in 48 patients (28%), while overall an increase in VDI score was recorded 65 times (IR: 8.1/100 PY). At the end of follow-up, the damage accumulation was increased, with 142 components (in total) occurring due to vasculitis-related damage, 41 due to GC toxicity and 28 due to comorbidities (online supplemental table 5). The Kaplan-Meier curve for the first VDI increase during maintenance is depicted in figure 3A. The risk for new damage accumulation by multivariable analysis was higher in those patients receiving GCs during maintenance (for every 5 mg increase in the average daily dose: HR=1.52, 95% CI 1.19 to 1.93, p=0.0007), those with persistently high disease activity during follow-up (BVAS at every visit: HR=1.42, 95% CI 1.25 to 1.61, p<0.0001) and older patients (for every 5-year increase: HR=1.10, 95% CI 1.03 to 1.18, p=0.0077, table 2).

Figure 3. Kaplan-Meier estimates of time up to the first: Damage accrual (A) and Hospitalisation (B) for the whole patient cohort. (A) Approximately 25% of VDI increase events take place in the first 3.3 years while median survival time is 9.1 years. (B) Approximately 25% of hospitalisations take place in the first 2.3 years while median survival time is 5.3 years. Censored data are noted with a cross on the Kaplan–Meier curve. ANCA, anti-neutrophil cytoplasmic antibody; AAV, ANCA-associated vasculitis; VDI, Vasculitis Damage Index.

Using the mixed-effects model for multiple events, only chronic GC use was associated with a higher risk for damage accumulation (for every 5 mg increase in the average daily dose: HR=1.14, p=0.033), while disease activity during follow-up or age were not (online supplemental table 6).

Hospitalisations: chronic GC use increases the risk of hospitalisations

Seventy-two (42.1%) patients required at least one hospitalisation during follow-up, while the total number of hospitalisations reached 132 (16.4/100 PY). 25% of hospitalisations were observed within 2.3 years from remission onset, as demonstrated by the Kaplan-Meier curve in figure 3B.

By multivariable analysis, chronic GC use (for every 5 mg increase in the average daily dose: HR=1.24, 95% CI 1.06 to 1.45, p=0.0060) and chronic damage (VDI at every visit, HR=1.32, 95% CI 1.10 to 1.57, p=0.0022) were associated with a higher risk (table 2). In the mixed-effects model for multiple hospitalisations, only chronic GC use was associated with a higher hospitalisation risk (HR=1.07, p<0.0001, online supplemental table 6).

Composite outcome: chronic GC use increases the risk of composite outcome

Using a composite outcome that included hospitalisation with VDI accumulation, 88 events were recorded. By multivariable analysis, older age (HR=1.22, 95% CI 1.03 to 1.19, p=0.0068), active disease both at diagnosis (HR 1.14, 95% CI 1.02 to 1.27, p=0.017) and during follow-up (HR=1.13, 95% CI 1.04 to 1.23, p=0.0039) and chronic GC use (HR=1.11, 95% CI 1.04 to 1.18, p=0.0008) were associated with a higher risk (table 2).

Sensitivity analysis

In a sensitivity analysis, using an IPTW-weighted MSM accounting for time-dependent confounding, GC exposure was not again significantly associated with the risk of major relapse (HR 0.82, 95% CI 0.35 to 1.91, p=0.64). Also, GC exposure remained significantly associated with VDI accumulation (HR 6.31, 95% CI 1.74 to 22.84, p=0.005). For hospitalisations, the association was in the same direction as in the primary analysis but did not reach statistical significance (HR 1.69, 95% CI 0.71 to 4.01, p=0.24). Overall, the findings remained directionally consistent with the primary analyses (online supplemental table 7).

Discussion

GCs have been traditionally used in addition to CYC and/or RTX for induction of remission in patients with AAVs. However, their role during maintenance is controversial. Our long-term, real life study clearly shows that chronic GC use, either continuously or intermittently, during the maintenance period, does not decrease the risk for major relapses while at the same time is associated with a higher risk for damage accumulation and hospitalisations.

In our longitudinal cohort study of 171 patients with AAV who had initially achieved remission either with CYC and/or RTX and subsequently received maintenance therapy, with or without GCs, we attempted to identify factors which were associated with major relapses. Both by multivariable time-varying and a mixed-effects model analysis, we found that RTX use was associated with a lower risk for major relapses. These findings are in agreement with findings from RCTs⁹ and real life¹⁰,¹² studies as well as with the most recent EULAR recommendations and ACR guidelines, which favour RTX as the preferred agent for maintenance of remission in patients with AAVs (GPA/MPA).^{3 4}

Data regarding the role of long-term GC use in preventing relapses are limited and rather contradictory. In an older meta-analysis that examined 13 studies performed between 1995 and 2007, the relapse rate in the 3 studies with patients who discontinued GCs was 14% compared with 43% in those who continued them (10 studies).¹⁴ However, it should be noted that most patients received maintenance therapy with AZA or methotrexate while there was insufficient information on severe relapses. In contrast, a retrospective study of 147 patients showed that those who had received GCs for more than 6 months had the same relapse risk with those who had stopped them.¹⁵ Nevertheless, detailed data regarding the dosage and duration of GCs were not given.

In our study, we recorded meticulously the patterns of GC use during the maintenance period. We reported that among those who had received GCs (87%), the majority were receiving them intermittently (~64%) rather than continuously (~36%). In both the multivariable analysis for the first major relapse and the mixed-effects model for multiple relapses, GC exposure during maintenance was not significantly associated with a lower risk of relapse, irrespective of the maintenance treatment used (RTX or non-RTX agents). Importantly, this finding remained directionally consistent in sensitivity analyses, which were performed to further account for potential time-dependent confounding related to evolving disease activity and treatment decisions during follow-up. Although GC continuation in clinical practice is partly influenced by physician assessment and perceived relapse risk, maintenance GC use in the present cohort also reflected the existing uncertainty regarding the optimal timing of GC withdrawal after remission, given the limited available evidence in this setting.

This is one of the few real life studies in the literature that addressed the role of long-term low-dose GCs during the maintenance period in preventing major relapses in patients with AAVs. While the results of The Assessment of Prednisone In Remission (TAPIR) randomised trial are awaited,¹⁷ our real life findings suggest that, within the limitations of an observational study, GCs during maintenance are not associated with an observable reduction in major relapse risk, after the first remission has been achieved.

Concerning damage accrual during follow-up, it was multifactorial reflecting the cumulative effects of prior vasculitis activity, GC exposure and comorbidities with GC-related toxicity. Despite the multifactorial nature of damage, GC-related toxicity contributed a significant proportion, mainly through musculoskeletal, ocular and metabolic domains. Therefore, a novel finding from our study was that chronic GC use was associated with increased risk of further damage accrual during the long-term follow-up. This was apparent with both statistical models used, while the sensitivity analysis yielded findings in the same direction. In the Wegener's Granulomatosis –Entretien trial (WEGENT) and the European Vasculitis Study Group trials, use of GCs at the end of the maintenance period was associated with an increased damage accrual (OR 1.12) or an increase in the VDI Score ≥5 (OR 1.26 per 12 months), respectively.^{18 19} Similarly, in a recent nationwide prospective study from Japan in patients with AAV and rapidly progressive glomerulonephritis, prednisone use at month 24 was associated with increased damage accrual between months 24 and 48 (OR 1.13).²⁰

Although previous studies have shown an increased risk for hospitalisations,²¹ mainly due to infectious causes,^{22 23} no other studies have looked thoroughly at the effect of chronic GC use in hospitalisation risk. In our study, during the approximately 7-year follow-up period, ~40% of patients needed to be hospitalised at least once (16.4/100 PY). Chronic GC use was identified as the only risk factor for hospitalisations by both statistical analyses, while chronic damage was associated with a higher risk only by the multivariable analysis for the first hospitalisation. The sensitivity analysis concerning hospitalisations presented results in the same direction as the main analysis.

The detrimental effects of chronic low-dose GC administration were also evident when the risk for the composite outcome of hospitalisations and damage accrual was estimated. Apart from chronic GC use, older age and persistently active disease (at diagnosis and follow-up) were also risk factors.

Strengths of our study include its real life, long-term (approximately 7 years), longitudinal design and the meticulous recording of the pattern of GC use (continuous, intermittent or no use) during the maintenance period. More specifically, it is important to note that the data were collected longitudinally and allowed the incorporation of time-varying variables, as well as the analysis of recurrent events for individuals who experienced the outcome more than once. The use of repeated measurements enables the monitoring of variable changes over time, taking into account that the therapeutic and disease course of most patients may involve multiple changes. Thus, its longitudinal design provided more robust data compared with those which could have been obtained from a purely retrospective study. Second, this was a large study of patients with AAV (n=171) followed in referral academic centres with expertise in AAV where similar therapeutic protocols for induction and maintenance of remission were employed. Furthermore, sensitivity analysis using MSM/IPTW models were performed for the main outcomes of the study in order to further address potential time-dependent confounding related to evolving disease activity and treatment decisions.

Our study also has limitations. Due to its observational design, causal inferences cannot be fully established, and residual confounding by indication may still be present despite adjustment and sensitivity analysis. Also, no standardised GC tapering protocol was applied across centres and physicians, reflecting real world clinical practice. Furthermore, although hospitalisations were restricted to clinically relevant admissions occurring in the context of AAV, more detailed classification of hospitalisation causes was not uniformly available for all admissions, and therefore hospitalisations were analysed as a broader clinically relevant outcome reflecting overall morbidity during maintenance follow-up. Finally, although the main longitudinal variables used in the analysis were systematically recorded during follow-up, missing or incompletely captured information inherent to retrospective data collection may have influenced the results.

In conclusion, while the optimal use of GCs during the induction phase of AAV management has been well established with a number of studies showing that lower dosing schemes and, in certain cases, use of novel agents such as the oral C5a inhibitor (avacopan),³²⁴,²⁶ may limit patients’ harmful exposure to GCs, their role during maintenance remains unclear. While the results of ongoing randomised trials looking at the use of GCs and of GC-sparing agents like avacopan beyond the initial period of remission are expected, our real life data suggest that low-dose GCs, given either continuously or intermittently during the maintenance period, were not significantly associated with a lower risk of major relapse. In contrast, GC exposure was associated with greater damage accrual and increased hospitalisation rates during follow-up.

Although our findings should be interpreted cautiously given the observational nature of our study and the potential for residual confounding, they suggest that higher GC exposure during maintenance therapy is not significantly associated with a lower risk for major relapse, while it is associated with higher risk for adverse outcomes. These results are informative for clinical decision-making regarding GC use in patients with GPA/MPA during the maintenance phase.

Data availability statement

Data are available upon reasonable request.