ANCA-associated vasculitis in Ireland: a multi-centre national cohort study

Introduction

Antineutrophil cytoplasmic antibody (ANCA)-associated vasculitis (AAV) is a rare multisystem autoimmune disease, characterised by necrotising pauci-immune small-vessel vasculitis. It comprises three clinico-pathological syndromes: granulomatosis with polyangiitis (GPA), microscopic polyangiitis (MPA) and eosinophilic granulomatosis with polyangiitis (EGPA). The estimated overall annual incidence of AAV is 13–20 cases/million¹ with a prevalence of 300–421 cases per million².

Significant treatment advances have occurred over the past 70 years with the advent of corticosteroid use in the 1950s, followed by the addition of cyclophosphamide in the 1960s, resulting in an improvement in two-year survival from 20 to 80%³. The introduction of ANCA testing in the 1980s improved detection of AAV and augmented disease awareness. With the achievement of remission in the majority, accompanied by an improvement in immediate survival, focus has now shifted to examining long-term outcomes, influenced by both treatment and the disease itself. AAV still carries a 2.7-fold increased risk of death compared to the general population⁴. Immunosuppression used to induce and maintain remission is a double-edged sword with >80% experiencing adverse events⁵. Given the relapsing and remitting nature of AAV, short randomised controlled trials (RCTs) are unable to provide robust evidence on longer-term outcomes, such as long-term treatment safety and efficacy in the biologic era. Additionally, RCTs do not provide information on diverse populations who may not consent to inclusion in trials, such as those with very severe disease admitted to ICU at presentation, older patients, marginalised populations and pregnant women. This knowledge gap has prompted the development of longitudinal vasculitis registries, including at least eight in Europe⁶. Increasingly, patient registries, linked to biobanks, are being used to facilitate longitudinal cohort studies in rare diseases, with a recent European effort to ensure standardisation and interoperability, via initiatives from FAIRVASC and the European Vasculitis Society (EUVAS) registries group⁶. This will allow collaborative research using aggregated data to analyse critical outcomes at a larger scale with sufficient power.

The Irish Rare Kidney Disease (RKD) registry and biobank was created in collaboration with the Irish national vasculitis patient organisation (https://vasculitis-ia.org/) in 2012 with the aim of collating clinical data and bio-samples from patients with vasculitis, and from disease and healthy controls on a national level. Importantly, this registry included, from inception, input from patient advocates (Vasculitis Ireland Awareness), recruitment during acute hospital admissions (allowing for pre-treatment data and sampling) as well as the involvement of a range of specialists including nephrology, rheumatology and immunology, allowing recruitment of broader clinical phenotypes. The current study arose from a need to appraise current practice and outcomes in the real-world setting. It aims to describe the long-term outcomes and potential prognostic factors at presentation in this large multicentre prospectively recruited cohort of AAV patients in Ireland.

Methods

Results

Patient survival in the first year from diagnosis (prospective AAV cohort)

In the first-year post diagnosis, twenty-two (5.5%) patients died, and actuarial survival was 93.8% (95% CI 91.3–96.3, Figure 1a). Infection was the leading cause of death (45.5%), followed by active vasculitis (18.2%, Extended data: Supplementary Table 4⁷). Age, degree of renal impairment at diagnosis, weighted infection and leucopenia adverse event scores, and cardiovascular involvement were independent predictors of one-year mortality on multivariate analysis (Table 2). An eGFR cut-off of <20ml/min/1.73m² identified those at highest risk of one-year mortality (Figure 1b)¹⁸. As the CBOE score rose, mortality increased. The optimal CBOE score (Extended data: Supplementary Figure 3)⁷ mortality cut-off was >8 (Extended data: Supplementary Figure 4)⁷, which was associated with a one-year survival probability of 55.9% (41.0–76.2) compared to 97.5% (95.8–99.2) for those with a score ≤8 (Figure 1c). Renal dysfunction severity was an independent risk factor for a CBOE score >8 (Table 3).

Figure 1. Factors associated with one-year mortality on univariate Kaplan-Meier analysis and compared using the log-rank test.

Optimal cut-points were obtained for each factor using the survminer package in R studio. one-year mortality a). overall and by b). estimated Glomerular Filtration Rate (CKD-EPI, eGFR) <= / >20 ml/min/1.73 m2, and c). Combined Burden of Events score (CBOE) <=/>8.

Table 2. Uni- and Multivariate Cox Proportional Hazards Models to investigate the factors associated with mortality within the first year from diagnosis.

Multivariate analysis was performed using backwards elimination. There was no co-linearity (>0.7) observed between the predictive factors.

AAV = ANCA associated vasculitis, GPA = granulomatosis with polyangiitis, PR3 = Proteinase-3 ANCA, eGFR = estimated Glomerular Filtration Rate (CKD-EPI), ml/min/1.73 m2, BVAS = Birmingham Vasculitis Activity Score, CYC = Cyclophosphamide, RTX = Rituximab, ENT = Ear, Nose and Throat, 95% CI = 95% Confidence Interval

	Univariate		Multivariate
Explanatory variable	Hazard ratio (95% CI)	p Value	Hazard ratio (95% CI)	p Value
Age at diagnosis, years	1.09 (1.05–1.14)	<0.001	1.07 (1.02–1.12)	0.009
Female (Ref: Male)	1.79 (0.77–4.14)	0.174
AAV Phenotype: Not GPA (Ref: GPA)	1.27 (0.52–3.10)	0.607
ANCA: Not PR3 (Ref: PR3-ANCA)	1.67 (0.68–4.10)	0.262
Induction treatment (Ref: CYC)
RTX	1.42 (0.51–3.99)	0.506
CYC & RTX	1.56 (0.34–7.20)	0.571
Other	2.30 (0.77–6.86)	0.136
eGFR at diagnosis	0.97 (0.95–0.99)	0.013	0.97 (0.94–0.99)	0.026
Weighted infection burden score	1.16 (1.11–1.21)	<0.001	1.11 (1.05–1.18)	<0.001
Weighted leucopenia burden score	1.37 (1.21–1.54)	<0.001	1.27 (1.11–1.44)	<0.001
Weighted adverse event (other) burden score	1.11 (1.07–1.15)	<0.001	1.05 (1.00–1.11)	0.07
Organ Involvement
Lung	2.16 (0.88–5.31)	0.092
Nervous	0.85 (0.20–3.65)	0.832
ENT	0.71 (0.28–1.81)	0.474
Cardiovascular	3.55 (0.83–15.18)	0.088	20.71 (3.84–111.81)	<0.001
Observations	397		395
Concordance			0.91

Table 3. Multivariate logistic regression to identify factors associated with a combined burden of adverse events (CBOE) score greater than 8, within the first year from diagnosis.

There was no co-linearity (>0.7) observed between the predictive factors. N = 390 as one participant had missing age and six were missing induction treatment data. AAV = ANCA associated vasculitis, GPA = granulomatosis with polyangiitis, eGFR = estimated Glomerular Filtration Rate (CKD-EPI), ml/min/1.73 m2, 95% CI = 95% Confidence Interval

Explanatory variable	Odds ratio (95% CI)	p Value
Age at diagnosis, years	1.03 (1.00–1.06)	0.065
Female (Ref: Male)	1.10 (0.55–2.21)	0.78
AAV Phenotype: Not GPA (Ref: GPA)	0.95 (0.45–2.15)	0.90
eGFR at diagnosis	0.98 (0.96–0.99)	0.01

Overall patient survival (prospective AAV cohort)

During the study period 55 patients (13.9%) died. Two- and five-year patient survival was 90.7% (87.7–93.9) and 76.8% (70.8–83.3), respectively (Figure 2a). Figure 2b and Extended data: Supplementary Table 5⁷ summarise the factors associated with overall mortality. After controlling for age and other potential confounders, only eGFR at diagnosis and CBOE score were independently significantly associated with mortality. The CBOE mortality association was primarily driven by the cumulative infection score.

Figure 2.

a). Kaplan-Meier curve demonstrating overall patient survival (solid line) with 95% CI (dashed lines), b). Forrest plot depicting the findings of a multivariate Cox Proportional Hazards Model to investigate the factors associated with overall mortality.

Renal survival (prospective AAV cohort)

End-stage kidney disease (ESKD) developed in 73 patients (18.4%). 1-, 2- and 5-year actuarial renal survival was 85.1% (81.5–88.8), 83.3% (79.6–87.3) and 79.2% (74.4–84.4), respectively (Figure 3a). ESKD-free patient survival at 1-, 2-, and 5-years was 80.6% (76.6–84.7), 77.7% (73.5–82.1) and 63.9% (57.9–70.5), respectively (Figure 3b). At initial presentation 84 (21.1%) required renal replacement therapy, of whom 37 (44.4%) subsequently became dialysis independent. Of those who had renal involvement (86%), the median eGFR at diagnosis was 22 ml/min (IQR 11–42, Extended data: Supplementary Figure 5)⁷. Factors independently associated with ESKD risk included degree of renal impairment and having >50% globally sclerosed glomeruli on kidney biopsy (Table 4, model 2). usCD163 was measured at diagnosis in a subgroup of 103 patients, in whom usCD163 was tightly associated with ESKD, even after adjustment for eGFR (Table 4, model 3). The optimal diagnosis eGFR and usCD163 cut-points which maximally split the cohort into low and high-risk groups of ESKD were 12ml/min/1.73m² and 740ng/mmol respectively (Figure 3c–d and Extended data: Supplementary Figure 6a–b)⁷. Patients with a ‘high’ (>740 ng/mmol) baseline usCD163 concentration were over five times more likely to develop ESKD, after adjusting for key confounders (Extended data: Supplementary Figure 7)⁷.

Figure 3. Factors associated with end-stage kidney disease (ESKD) on univariate Kaplan-Meier analysis and compared using the log-rank test.

Optimal cut-points were obtained using the survminer package in R. a). Overall renal survival, b). ESKD-free patient survival, and renal survival stratified by c). estimated Glomerular Filtration Rate (CKD-EPI, eGFR) at diagnosis of <= / >12 ml/min/1.73 m2 and d). urine soluble CD163 at diagnosis (usCD163) of <= / >740.4 ng/mmol.

Table 4. Uni- and Multivariate Cox Proportional Hazards Models to investigate the factors associated with end-stage kidney disease.

^aModel 1 included baseline eGFR, adjusted for age at diagnosis, gender, AAV phenotype, ANCA serotype and the occurrence of a renal relapse (N = 397).

^bModel 2 included baseline eGFR, adjusted for age at diagnosis, gender, AAV phenotype, ANCA serotype and Berden score on diagnostic renal biopsy (N = 244 as 153 participants did not have biopsy data).

^cModel 3 included baseline uCD163, adjusted for baseline eGFR, age at diagnosis, gender, AAV phenotype and ANCA serotype (N = 103 as 294 participants did not have baseline uCD163 data).

There was no co-linearity (>0.7) observed between the predictive factors included in the multivariate models. *Missing data: uCD163 N = 294, Berden score N = 153. AAV = ANCA associated vasculitis, GPA = granulomatosis with polyangiitis, PR3 = Proteinase-3 ANCA, eGFR = estimated Glomerular Filtration Rate (CKD-EPI), ml/min/1.73 m2, BVAS = Birmingham Vasculitis Activity Score, RRT = Renal Replacement Therapy, 95% CI = 95% Confidence Interval.

	Hazard ratio (95% CI, p Value)
	Univariate		Multivariate
Explanatory variable		Model 1a	Model 2b	Model 3c
Age at diagnosis, years	1.02 (1.00–1.04, p=0.054)	0.996 (0.98–1.02, p=0.708)	0.996 (0.97–1.02, p=0.709)	1.01 (0.98–1.05, p=0.498)
Female (Ref: Male)	0.68 (0.42–1.11, p=0.126)	0.66 (0.40–1.10, p=0.110)	0.65 (0.37–1.15, p=0.141)	0.17 (0.05–0.62, p=0.007)
AAV Phenotype: Not GPA (Ref: GPA)	2.78 (1.55–4.98, p=0.001)	1.34 (0.62–2.91, p=0.456)	0.94 (0.41–2.14, p=0.885)	1.28 (0.30–5.38, p=0.739)
ANCA: Not PR3 (Ref: PR3-ANCA)	1.87 (1.13–3.09, p=0.014)	0.86 (0.45–1.64, p=0.654)	0.69 (0.35–1.35, p=0.275)	2.09 (0.58–7.57, p=0.263)
eGFR at diagnosis	0.92 (0.89–0.94, p<0.001)	0.92 (0.89–0.94, p<0.001)	0.93 (0.90–0.96, p<0.001)	0.97 (0.94–0.995, p=0.02)
Required RRT at initial presentation	13.26 (8.01–21.96, p<0.001)	-	-	-
Renal relapse	0.74 (0.30–1.84, p=0.518)	0.57 (0.23–1.42, P=0.225)	-	-
Berden score on diagnostic renal biopsy (Ref: Focal)
Crescentic	3.10 (1.16–8.32, p=0.024)	-	0.80 (0.27–2.36, p=0.683)	-
Mixed	3.62 (1.42–9.19, p=0.007)	-	2.13 (0.95–4.79, p=0.275)	-
Sclerotic	12.14 (4.93–29.86, p<0.001)		5.76 (2.61–12.71, p=0.001)	-
uCD163 at diagnosis (µg/mmol)	1.37 (1.13–1.68, p=0.002)	-	-	1.76 (1.24–2.50, p=0.002)
Observations	397*	397	244	103
Concordance		0.87	0.84	0.80

Discussion

We describe the Rare Kidney Disease (RKD) national Irish registry at the 10^th anniversary of its inception. This multicentre longitudinal cohort study aimed to characterise the Irish AAV cohort, describe their long-term outcomes and identify baseline predictors of these outcomes. The cohort comprises an ethnically homogeneous Caucasian population, with most displaying renal involvement and hence a slight MPA and MPO-ANCA predominance was observed. This contrasts to the GPA preponderance^20–23 in most other European registries, which tend to have lower rates of renal involvement. Our cohort were typically older, in agreement with recent studies showing an older age at diagnosis for those with MPA^21,24,25. We noted a slight male preponderance, consistent with a prior meta-analysis of EUVAS RCTs²², which differs from the equal gender split most commonly reported^2,20,21,23). Notably, the diagnostic delay (median 9.4 weeks, IQR 4.3–25.6, N = 238) in our cohort was somewhat shorter than other reported series (median range 12–52 weeks^21,26–29).

Patient survival in our real-world cohort was comparable to the international literature, predominantly based on RCTs, which ranges from 80–97%, 80–85% and 70–80% for one, two and five years respectively^{2,5,22,26,30,31}). One may have expected lower rates given our relatively older cohort with frequent renal insufficiency. We too noted a steep decline in survival in the initial 90 days, when maximal concomitant disease activity and immunosuppression occurs. Infection was the primary cause of death, consistent with previous reports^{5,22,26,32,33}). Multiple studies have demonstrated an inverse relationship between survival and the number of accrued infections, particularly in the first year^31,34). In our study, the weighted infection and leucopenia scores, reflecting the accumulation of these adverse events, were the strongest predictors of one-year mortality. This highlights the growing need for personalisation of immunosuppression to simultaneously achieve disease control while avoiding unnecessary excess immunosuppression, in addition to steroid-sparing, borne out by the PEXIVAS trial³⁵. It also serves as a key reminder to clinicians to be alert to early infections and to routinely assess the need for infection prophylaxis against Pneumocystis jirovecii (PCP), fungi and viruses, as well as timely influenza, pneumococcal, hepatitis B virus and SARS-CoV-2 vaccination, as advised in guidelines³⁶. Notably, only one patient died from PCP in our cohort, which may signify the success of trimethoprim/sulfamethoxazole prophylaxis, which is cost-effective in AAV³⁷.

Our study supports the previously observed association between increasing age, impaired kidney function and early mortality^{5,20,21,30,33,38–41}). While most studies have previously attributed this to excess severe infection^{21,31,32,42,43}, we observed that advanced age and renal dysfunction are independent predictors, after adjustment for adverse events. Renal impairment results in immune dysfunction⁴⁴ and altered pharmacokinetics, limiting drug clearance and hence increasing the risk of infection and treatment toxicity⁴⁵. Our findings further support the use of dosing nomograms for cyclophosphamide based on age and eGFR to reduce treatment-induced leucopenia, thereby reducing mortality^32,45. Future prospective research should explore whether a similar approach is necessary for rituximab, rather than the current fixed or weight-based dosing regimens.

Our study validates the use of the combined burden of events (CBOE) score⁵, to predict early mortality in an independent real-world cohort. Building on the initial description, we identified 8 as an optimal cut-off score to stratify patients into low/high risk of early death. Like the original description, a score ≤8 was associated with <5% risk of death within the first year. This risk rose 16-fold with a score >8. Analogous to scoring BVAS and Vasculitis Damage Index (VDI) during routine follow-up, clinicians should consider monitoring the cumulative CBOE score. A rising score warrants active assessment and implementation of strategies to reduce further toxicity. We also confirmed that renal dysfunction severity is the strongest independent risk factor of the CBOE score, underpinning the critical importance of accounting for kidney function when prescribing immunosuppression⁵. We found no statistical difference in the risk of a high CBOE score between the cyclophosphamide and rituximab groups, in keeping with the findings of the RAVE trial⁴⁶. Our findings emphasise that no induction regime is ‘safe’ and future research should focus on developing remission induction strategies that minimise toxicity, while maintaining similar efficacy.

Reported renal survival in AAV is variable, largely related to differing baseline patient and disease characteristics across cohorts. Our one-, two- and five-year renal survival estimates of 85.1%, 83.3% and 79.2% respectively are in line with other series^{26,31,47–50}): 20–40% of AAV patients reach ESKD by 5–7 years. Aligned with our relatively short diagnostic delay, the baseline renal function (median eGFR 22 ml/min 1.73m²) in those with renal involvement was comparable to international benchmarks (median eGFR range 12–25 ml/min/1.73m²)^51,52. Our results confirm the importance of baseline renal dysfunction^{26,31,32,47,49,53} and of histological evidence of renal scarring⁵⁴ in determining ESKD risk. Contrary to prior series, we did not find an association between ANCA serotype⁵⁰ or prior relapse⁴⁷ and ESKD, in multivariate analysis. Our group has recently demonstrated the benefit of usCD163, a glomerular macrophage marker, as a non-invasive tool for the diagnosis of AAV renal relapse⁵⁵. Another group highlighted its use in the prediction of doubling of serum creatinine in lupus nephritis⁵⁶. We demonstrate, for the first time, the potential use of baseline usCD163 in predicting ESKD at diagnosis, even after adjustment for renal insufficiency. We identified a cut-off of 740ng/mmol to stratify patients into low/high risk groups. The risk of ESKD was five-times higher for participants with a baseline usCD163 of >740.4ng/mmol compared with those below this cut-point. Once validated in independent cohorts, these findings will contribute to the personalisation of immunosuppression to minimise ESKD risk – a costly outcome to patients and society.

The main strengths of our study include the well-characterised national AAV cohort, recruited using uniform classification criteria. Our cohort does not exclude patients with severe disease, EGPA nor older people – commonly excluded in RCTs – and thus serves as a source of real-world data, which is important as outcomes differ to those reported from RCTs⁵⁷. The use of REDCap as the registry infrastructure, with standardised interoperable ‘instruments’ to enable uniform data collection is another key strength. Our registry formed the basis of the model EUVAS registry which is being applied in nascent vasculitis registries across Europe. This capability is in line with both the EUVAS⁶ and European Reference Network initiatives (http://rita.ern-net.eu/about-rita/mission-goals-and-objectives/) to align and integrate European registries, with the aim of improving both care and outcomes in rare disease.

We also acknowledge the limitations of our study. As with any observational research, missing data is always a challenge. For this reason, we were unable to include BVAS in multivariate models. Although all relevant specialties were represented, most participants were recruited through nephrology centres, reflected in the high degree of renal involvement, which limits the generalisability of our findings to non-renal populations. The strengths and limitations of CBOE score were previously discussed in detail⁵. Additionally, the CBOE score was not adjusted for prophylaxis use. Induction treatment was included as a categorical variable – the variation in intensity of same (denoted by cumulative dosing) was not available. Going forward, we aim to capture the intensity of induction immunosuppression in our registry.

Conclusions

In conclusion, we demonstrated that long-term outcomes from a real-world Irish cohort are similar to international standards. Our results highlight that despite improvement in outcomes over the last number of decades, treatment toxicity is an ongoing concern and individualisation of immunosuppression relative to disease severity and relapse risk remains a key unmet need. Our findings serve to remind clinicians that the accumulation of adverse events, both in the first year and beyond, particularly in older patients and those with renal insufficiency, is strongly associated with premature death and should be accompanied by increased monitoring and adjustment of immunosuppression to further limit treatment-related morbidity. The degree of renal impairment at presentation is also a key determinant in renal survival, hence early identification remains a critical goal. Baseline usCD163 shows promise as a biomarker for ESKD prediction. Future research should further explore this in the context of a multi-modal approach to personalisation of AAV treatment.

Consent

Written informed consent for publication of the patients’ details was obtained from the patients.