Background/Purpose: Previous analyses of the FVSG- and Pisa-cohort EGPA patients identified several covariates associated with poor outcomes and suggested differences between ANCA+ and ANCA– EGPA-patient subsets. Reanalysis of the FVSG cohort using more stringent (than the 1990 ACR criteria) EGPA definitions and clustering analysis identified 3 main clusters reinforcing the impact of ANCA status. We compared the FVSG and Pisa cohorts and aimed to cross-validate the FVSG-cohort cluster analysis results with Pisa-cohort data.

Methods: We first compared the main characteristics of patients from both cohorts, selecting only those patients with definite EGPA (biopsy-proven vasculitis, ANCA+ and/or surrogate clinical manifestations of vasculitis, i.e. mononeuritis multiplex, purpura, alveolar hemorrhage and/or renal disease [glomerulonephritis and/or hematuria]) and known ANCA status. Hierarchical cluster analysis of FVSG-cohort patients included 6 clustering parameters (cutaneous manifestations, sinusitis, cardiomyopathy, renal disease, mononeuritis multiplex, ANCA) and yielded 3 main clusters, further characterized using K-means nonhierarchical clustering methodology. The same cluster analysis was applied to the Pisa patients.

Results: Table 1 summarizes the main characteristics of both EGPA-patient cohorts with biopsy-proven vasculitis, ANCA+ and/or surrogate clinical vasculitis manifestations.

	Pisa patients N = 51	FVSG patients N = 241	P
Sex, (M/F, n)	24/27	132/109	0.32
Age at diagnosis (yr)	52	51	0.55
Asthma	100%	96%	0.14
ENT manifestations	86%	55%	<0.01
Lung infiltrate	45%	44%	0.84
Alveolar hemorrhage	2%	6%	0.22
Skin manifestations	61%	49%	0.13
Purpura	49%	33%	0.02
Mononeuritis multiplex	61%	64%	0.72
CNS	4%	6%	0.52
Cardiomyopathy	12%	18%	0.26
Severe GI signs	8%	6%	0.58
Eye	12%	8%	0.37
Renal disease (n with GN)	6% (2)	21% (12)	0.01
BVAS at diagnosis	17.7±7.7	20.8±8.1	0.02
ANCA+	57%	45%	0.12
Vasculitis relapse	27%	32%	0.57
Deaths	6%	12%	0.18

Cluster analysis identified 3 clusters in the Pisa cohort:  

• n=21, all ANCA+, with 86% ENT manifestations, 81% peripheral nerve involvement, 48% skin disease, 14%cardiomyopathy, 10% renal disease, 38% vasculitis-relapse rate, 12% died;
• n=14, all ANCA–, with 100% with peripheral nerve involvement, 72% ENT manifestations, 36% skin involvement, 7% cardiomyopathy, 7% renal disease, 29% relapsed, none died;
• n=16, 50% ANCA+, and 100% with ENT, 100% skin signs, 13% cardiomyopathy, but no renal disease and no peripheral nerve involvement, 13% relapsed, 7% died.

As for the FVSG cohort, the main feature distinguishing between clusters was ANCA status. Although, the intermediate ANCA+ (low but not null %) clusters in both cohorts had the lowest frequencies of renal disease, highest of skin disease and intermediate cardiomyopathy, distributions of other EGPA manifestations among clusters differed between the 2 cohorts.  

Conclusion: Cluster analysis of 2 independent EGPA-patient cohorts demonstrated that ANCA status is the main parameter distinguishing disease subsets, as previously found with other statistical methods. These exploratory analyses exemplify the advantages and challenges of cluster-analysis methodology. Interpretation of results must be cautious, taking into account the methods used and sample size.

---

« Back to 2013 ACR/ARHP Annual Meeting

ACR Meeting Abstracts - https://acrabstracts.org/abstract/eosinophilic-granulomatosis-with-polyangiitis-churg-strauss-syndrome-comparison-of-the-independent-french-vasculitis-study-group-and-italian-pisa-patient-cohorts-and-cross-validatio/