Clinical Characteristics, Predictors for Mortality and Comparison of the Birmingham Vasculitis Activity Score and the Five-Factor Score on Survival in ANCA-Associated Vasculitis in Hong Kong

INTRODUCTION

Antineutrophil cytoplasmic antibody (ANCA)- associated vasculitis (AAV) is a distinct group of systemic vasculitis with multi-organ involvement and varying clinical presentation. Three main subtypes have been described in the literature, namely granulomatosis with polyangiitis (GPA), microscopic polyangiitis (MPA), and eosinophilic granulomatosis with polyangiitis (EGPA) [1, 2]. The incidence and prevalence of each AAV vary in different localities. Studies from China and Japan suggested MPA being the most common form of AAV, with a higher prevalence of myeloperoxidase (MPO) AAV, whereas in the Western counterpart, proteinase-3 (PR3) ANCA GPA is more prevalent [2, 3, 4, 5, 6, 7]. Published local data on the clinical characteristics and outcomes of patients with AAV in Hong Kong are scarce.

The revised Five-Factor Score in 2009 (FFS-2009) (Supplementary Figure 1A) is recommended in international guidelines for initial treatment decisions [8, 9, 10]. The Birmingham Vasculitis Activity Score (BVAS) (Supplementary Figure 1B) is a validated activity scoring system frequently used in therapeutic studies of systemic vasculitis to document disease activity. A previous study suggested that higher BVAS at baseline confers a greater mortality risk [11]. Previous overseas registry suggested both FFS-2009 and BVAS were independent factors related to mortality, with the 2009 FFS showing better accuracy in predicting mortality upon initial diagnosis [12].

The aim of this study was to describe the clinical profile and predictors of mortality of AAV patients managed in the three regional hospitals (Caritas Medical Centre, Princess Margaret Hospital, and Yan Chai Hospital) in Hong Kong. The performances of FFS-2009 and BVAS in predicting mortality among AAV patients were reviewed with this local cohort.

MATERIALS AND METHODS

This was a multicenter, retrospective cohort study. Data were extracted from the hospital authority (HA) electronic system to identify patients with AAV being managed in the three local hospitals from January 1, 2011 to March 31, 2022. Patients with diagnosis codes of MPA [446.0(2)], necrotizing vasculopathy [446.0(7), 446.0(8)], polyarteritis [446.0(9)], Wegener’s granulomatosis [446.4], Churg–Strauss syndrome with glomerulonephritis [710.8(1)], and arteritis unspecified [447.6], according to the ninth revision of the International Classification of Diseases (ICD-9) were reviewed. Patients were included if they were adults aged 18 or above at diagnosis and fulfilled the American College of Rheumatology (ACR) criteria [13, 14, 15, 16] or the Chapel Hill Consensus Conference (CHCC) definition (1) of AAV, with the consideration of the European Medicines Agency (EMA) algorithm [17]. Drug-induced AAV and AAV overlapping with other connective tissue diseases were excluded.

Patient demographics, comorbidities, diagnosis type, date of diagnosis, ANCA specificities, clinical presentations, laboratory findings, histology results, treatment profile, and the disease course were reviewed. FFS-2009 and BVAS at initial diagnosis were calculated retrospectively for each included patient and analyzed.

Definition

The date of AAV diagnosis was defined as follows: start of treatment with prednisolone $\ge 30$mg per day or equivalent glucocorticoid; if untreated or missing data, the day of biopsy; and if no biopsy, the day of the first positive ANCA test result [18]. Proteinuria was measured by 24-hour urine protein or an equivalent spot urine protein-to-creatinine ratio. The clinical profile was documented according to BVAS and FFS items. Pulmonary-renal syndrome (PRS) was defined as an estimated glomerular filtration rate (eGFR) $<50$mL/min/1.73m², with evidence of microscopic hematuria and pulmonary hemorrhage [19]. Pulse steroid therapy was defined as the administration of intravenous methylprednisolone $\ge 500$mg daily, or equivalent, during initial treatment [8, 19]. Maintenance therapy was defined as administering non-glucocorticoid immunosuppressive agents after 6 months of initial induction therapy [20, 21].

Statistical analyses

Categorical data were reported as frequency (percentage) and compared using the Fisher’s exact test or the Pearson chi-square test. Continuous variables with a normal distribution were reported as the mean (standard deviation [SD]), and continuous variables with a non-normal distribution were reported as the median (interquartile range [IQR]). They were compared with parametric tests (student’s t-test or one- way ANOVA) or nonparametric tests (Mann–Whitney U test or Kruskal–Wallis test). Univariate analyses were performed to identify potential predictors of mortality. Clinically significant variables with $P<0.1$ in univariate analysis were entered into the multivariate Cox proportional hazards model. Survival analyses were conducted with the Kaplan–Meier method and compared using the log-rank test. The accuracy of BVAS and FFS in the prediction of mortality was analyzed with receiver operating characteristic (ROC) curves. Their respective areas under the curve (AUC) were compared using the DeLong test. The optimal cut-off value for BVAS in the prediction of mortality was determined by the Youden index. A P-value of $<0.05$ was considered statistically significant. Statistical analyses were performed by using the Statistics Package for Social Sciences (version 29.0; IBM, Armonk, NY, USA).

Ethical approval

The study was conducted in accordance with the Declaration of Helsinki. Consents were waived in view of the retrospective nature of the study. This study was approved by the KWC Research Ethics Committee (KW/EX-22-050(172-06)).

RESULTS

Demographics

Eighty-three AAV patients were included in the study for analysis (Supplementary Figure 2). The median follow- up duration was 23 months (IQR 3, 49). The baseline demographics at diagnosis are shown in Table 1. The median age at diagnosis was 70.5 (IQR 68.7, 78.2). MPA patients were significantly older at diagnosis, comprising more than two-thirds of the cohort. One-third of the EGPA cases were ANCA-negative. A significantly higher proportion of patients with EGPA had preexisting asthma at the time of initial diagnosis.

Clinical characteristics and laboratory features

The initial clinical manifestations are summarized in Tables 2A, B. Fewer MPA patients had ENT symptoms ($P<0.001$ for GPA and $P=0.097$ for EGPA). GPA patients had more eye and mucosal features ($P<0.001$ for both MPA and EGPA), as well as gastrointestinal features ($P=0.067$ for MPA, $P=0.017$ for EGPA). More EGPA patients had neurological involvement ($P<0.001$ for MPA, $P=0.007$ for GPA) and cutaneous manifestation at presentation. EGPA patients had significantly less renal manifestation, when compared with MPA ($P=0.029$) but not GPA ($P=0.125$). Laboratory findings are shown in Table 2C. The median peak serum creatinine and the eGFR were 229$\mu$mol/L (IQR 102, 553) and 19.5mL/min/1.73m² (IQR 8.55, 48.0), respectively. Twenty-four patients (28.9%) had eGFR $\le 10$mL/min/1.73m². MPA patients presented with significantly lower hemoglobin ($P=0.001$) and worse kidney function ($P<0.001$).

Treatment profile

Treatment profiles are summarized in Table 3. About two-thirds of the cohort received an intravenous pulse steroid. Apart from steroids, cyclophosphamide was the most frequently used induction agent. Nearly one-third of patients required temporary renal replacement therapy during the initial phase. Fifty (60.2%) patients received maintenance therapy with non-glucocorticoid agents at 6 months. Azathioprine was the most frequently administered first-line maintenance agent. About one-third of patients did not receive any prophylaxis for pneumocystis infection.

Mortality

Thirty-seven (44.6%) patients died in the study period. Nineteen (22.9%)patients died at the same index admission. Among those deceased, the median time to death was 95 days [IQR 29, 1302]. The 30-day, 1-year, 5-year mortality rates were 10.8%, 30.1%, and 38.6%, respectively (Table 4). In subgroup analysis, patients with EGPA had significantly longer follow-up periods, median survival, and time to mortality events (Figures 1A and 1B).

Predictors of mortality in the AAV cohort

Univariate and multivariate Cox regression analyses were performed (Tables 5A and 5B). In the multivariate Cox regression model, five clinical predictors were shown to be associated with overall mortality, including age (HR 1.043, for each additional year of age, $P=0.027$), stabilized peak serum creatinine (HR 1.002, for every unit increase, $P=0.001$), hemoglobin level (HR 0.754, for each unit decrease, $P=0.006$), cardiac involvement (HR 3.862, $P=0.008$), and the use of maintenance therapy (HR 0.261, $P=0.002$).

The performances of FFS-2009 and BVAS in prediction of survival among AAV patients

The FFS-2009 was significantly higher among the non-survivors when compared with the survivors ($P<0.001$). The BVAS was also higher among those deceased than those who lived ($20.0\pm 8.14$ vs. $16.95\pm 7.47$, respectively; $P=0.044$) (Table 6).

The overall survival of patients with FFS-2009 scores of 0, 1, and $\ge 2$ was 85.7%, 84.6%, 36%, respectively ($P<0.001$) (Figure 2A). When cases were stratified between $\mathrm{\text{FFS}}<2$ and FFS $\ge 2$, the overall survival rate was significantly lower among the group with higher FFS (84.8% vs. 36.0%, $P<0.001$, log-rank test $P<0.001$), as shown in Figure 2B. By univariate Cox regression, the hazard ratio of FFS $\ge 2$ was 8.056 ($P<0.001$, 95% CI [3.046, 21.307]).

ROC analysis of BVAS revealed an optimal cut- off value of 15.5 (sensitivity 88.2%, specificity 38.6%, Youden index 0.268). A cut-off score of 16 was chosen. The hazard ratio of BVAS $\ge 16$ was 2.99 ($P=0.024$, 95% CI [1.158, 7.721]). ROC analysis for the overall survival of the entire cohort yielded AUC values of 0.794 (95% CI [0.697, 0.892]) and 0.629 (95% CI [0.508, 0.750]) for FFS-2009 and BVAS, respectively, suggesting that FFS-2009 had better accuracy in predicting overall survival at the time of diagnosis in patients with AAV (AUC difference 0.165, 95% CI [0.038, 0.293], $P=0.011$) (Figure 3A).

Similar results were observed when we compared the performance of the FFS-2009 to that of BVAS in predicting our cohort’s 1-year and 5-year survival (Figures 3B and 3C). For 1-year survival, ROC analysis yielded AUC values of 0.874 (95% CI [0.800, 0.948]) and 0.700 (95% CI [0.577, 0.823]) for the FFS and the BVAS, respectively. For 5-year survival, ROC analysis yielded AUC values of 0.818 (95% CI [0.722, 0.913]) for FFS-2009 and 0.629 (95% CI [0.506, 0.753]) for BVAS.

DISCUSSION

Previous publications [4, 5, 6, 22, 23] suggested significant geographical variation regarding AAV patients’ demographic and baseline characteristics between the East and the West. In our cohort, MPA was the most common subtype of AAV, accounting for 70% of total cases. MPO-ANCA was tested positive in 81% of all cases. PR3-ANCA was detected almost exclusively in GPA only. These results echoed the findings from previous Asian studies [3, 7].

AAV usually occurs in the sixth decade of life [5]. MPA patients tend to be older when compared with other AAV subtypes. Our cohort’s median age at diagnosis was 71, which was numerically higher than that in overseas cohorts [5, 24, 25, 26, 27, 28, 29]. The median age at diagnosis of MPA in the current cohort was 74, significantly older than that of GPA and EGPA, with medians of 56 and 65, respectively. This finding was similar to reports from other parts of the world [5, 24, 25, 26, 27, 28, 29].

Each AAV subtype has a known predilection for different organ systems [5, 30]. Both MPA and GPA have frequent renal involvement. About 80% of MPA and GPA cases in our cohort showed acute renal impairment. In contrast, severe renal involvement was uncommon in EGPA. While 10%–20% of the MPA and GPA cases presented with PRS at diagnosis, none of the EGPA cases developed PRS or required temporary RRT. Cutaneous and peripheral nervous involvement were frequently observed in EGPA patients, up to 70% and over 90%, respectively. However, those features were less commonly reported in patients with other AAV subtypes. The abdominal, respiratory, and cardiac involvement frequencies were comparable to overseas cohorts [24, 25, 26]. In contrast, our cohort’s ENT, eye, and mucosal involvement frequencies (Table 2A) were relatively low compared to cohorts overseas, possibly due to lower proportions of GPA and EGPA cases. Further subgroup analysis on EGPA, based on the ANCA status, was impossible with the small sample size. A meta-analysis of 24 EGPA cohorts suggested apparent differences in the pattern of organ involvement between ANCA-positive and ANCA-negative EGPA. Peripheral neuropathy, renal involvement, and cutaneous purpura were more often seen in ANCA-positive EGPA; Pulmonary infiltrates and cardiac involvement were more frequent in ANCA-negative EGPA [31].

Our cohort may be represented by AAV patients with more severe and active disease at the initial presentation. Over 60% of our cases fell into the worst FFS category group (FFS-2009 $\ge 2$). In contrast with the cohort of the French Vasculitis Study Group, only about 15% of the entire cohort had FFS $\ge 2$; one-third of the cohort had FFS of 1; and around half of their cases had FFS of 0 [10].

In our cohort, half of the mortality occurred in the first admission for AAV, which is an alarming finding. The diseasese verity may partly explain the high mortality risk in the initial phase. Nevertheless, treatment-related complications like infection also contributed to the mortality. To reduce the high mortality, a high index of suspicion for potential cases, early establishment of diagnosis, prognostic stratification, and optimization of induction treatment, while balancing the risk of over-immunosuppression, are of utmost importance.

The 1-year and 5-year survival rates were 69.9% and 61.4%, respectively, in our cohort. The findings were comparable to an earlier systemic review in 2008, which reported the 5-year survival of AAV patients was 69%–91% for GPA, 45%–76% for MPA, and 60%–97% for EGPA [32]. In more recent studies, the 5-year survival rate of AAV patients improved to around 80%–90% [27, 28, 29, 33]. Such survival improvement was also demonstrated in a retrospective study by the French Vasculitis Study Group. It analyzed the mortality data of patients with systemic necrotizing vasculitis from their registry from before 1980 to after 2010. The 5-year survival rate improved from 72.2% (before 1980) to 94.5% (after 2010), with a decrease in mortality by three-fold [34].

The increasing availability of alternative induction agents, for example, rituximab and non- cyclophosphamide maintenance therapy, is expected to mitigate the dose-dependent cancer risk and toxicity due to the alkylating agent. The advocation of a reduced-dose glucocorticoid regimen and the use of PCP prophylaxis, supported by more recent clinical studies [19, 35] and international guidelines [8, 36], is expected to reduce the risk of infection due to heavy immunosuppression. These changes are going to improve the clinical outcomes of AAV patients.

In our study, we identified five independent clinical variables with significantly greater risk of death by multivariate Cox regression analysis: increasing age, higher peak serum creatinine, cardiovascular involvement, lower hemoglobin level, and the absence of non-glucocorticoid maintenance therapy. These independent risk factors were also reported in previous studies on mortality risk factors in patients with AAV [7, 24, 25, 26, 27, 28, 29, 33]. The first three factors were the components of the FFS, frequently used in daily practice for prognostication and decisions on the initial treatment strategy. Patients with severe disease manifestations may present with a lower initial hemoglobin level due to systemic inflammation, renal failure, or acute blood loss, for example, gastrointestinal bleeding and alveolar hemorrhage [33]. In the univariate Cox regression analysis, gastrointestinal bleeding and alveolar hemorrhage were identified as significant predictors of overall survival. The finding of neurological involvement as a protective factor in the univariate analysis should be interpreted with caution, as it was no longer significant after adjustment with age and serum creatinine level. ENT manifestation was included in the FFS as a protective factor. Previous studies on GPA and EGPA suggested the presence of ENT manifestations was associated with a better prognosis in terms of mortality [28, 29]. However, this association was not apparent in patients with MPA [27]. In our study, ANCA status was not shown to be a predictor of overall survival. The association between MPO-ANCA positivity and increased mortality risk was controversial. While some studies did not reveal any association between ANCA status and mortality [7, 24, 26, 27, 33], some suggested MPO-ANCA positivity, when compared with PR3- ANCA positivity, was associated with an increased risk of mortality [25, 28], possibly due to more frequent severe renal involvement in patients with positive MPO-ANCA.

Our results suggested that both FFS-2009 and BVAS at initial diagnosis can predict mortality. Both higher FFS (FFS $\ge 2$) and higher BVAS (BVAS $\ge 16$) were associated with an increased risk of mortality by up to 8-fold and 3-fold, respectively. FFS-2009 demonstrated better discriminating power in assessing early mortality events at 1 year (AUC $\mathrm{\text{value}}=0.874$) than late mortality at 5 years (AUC $\mathrm{\text{value}}=0.818$) and the overall survival (AUC $\mathrm{\text{value}}=0.794$). FFS-2009 performed better than BVAS (AUC values range from 0.629 to 0.700) in mortality prediction. One possible reason for the inferiority of BVAS in predicting mortality was that, despite the comprehensiveness of BVAS in evaluating AAV cases, some domains, like general, cutaneous, eye, andmucosal, werenotsignificantpredictorsofmortality. On the other hand, FFS included only five items, all with substantial prognostic impact. In clinical practice, FFS appears to be a better tool at initial assessment for its simplicity and reasonably good performance in predicting mortality. Yet, BVAS still plays a vital role in AAV assessment since it enables ongoing interval assessment of serial changes. For this reason, BVAS was recommended by EULAR in 2007 as the tool to define treatment response, remission, and relapse in AAV [37]. Our finding aligned with a recent study by the Spanish group [25], which included 550 AAV patients in its registry, analyzing the performance of FFS and BVAS in predicting survival. They concluded that FFS and BVAS were significantly related to death, with FFS-2009 being the stronger predictor with a hazard ratio of 2.9. Similarly, ROC analysis revealed FFS has better accuracy in the prediction of survival than BVAS. The AUC values for 1-year, 5-year, and overall survival were 0.80, 0.78, and 0.74 for FFS-2009, whereas for BVAS, they were 0.49, 0.56, and 0.60, respectively.

Limitations and future direction

Due to the retrospective and observational nature of this study, unstandardized data collection could lead to significant systematic errors. Subtle clinical manifestations might be overlooked and left unrecorded. The time of symptom onset was often unavailable in the record, making the assessment of diagnostic delay difficult. The date of diagnosis was retrospectively defined by the time of treatment initiation, commonly with at least moderate to high doses of glucocorticoid [18]. Since FFS and BVAS were backdated and missing data were unavoidable, the computed scores could be overestimated in FFS (e.g., ENT symptoms) or underestimated in BVAS. The number of GPA and EGPA cases was small in this cohort, which reduces its generalizability and limits further subgroup analyses due to a lack of power.

The current study included AAV patients from three centers in a retrospective manner. It provided insights into the clinical characteristics of AAV in this locality. A Hong Kong territory-wide study with a prospective design manner will allow a much more precise case definition, facilitate the collection of patient data, and enable a broader understanding of the local disease pattern, treatment profile, and prognosis. Establishing a local vasculitis registry will also allow better coordination and communication in handling cases among different rheumatology centers in Hong Kong.

The latest ACR AAV classification criteria in 2022 appear less complicated and easier to use for defining the cases for inclusion when compared with the old ones and the EMA algorithm [1, 14, 15, 16, 17]. However, further validation studies may be necessary before applying them locally.

In conclusion, despite the advances in treatment, AAV carries significant morbidities with high mortality, with 1-year and 5-year mortality rates of about 30% and 40%, respectively, in this cohort. Multivariate Cox regression analysis suggested several features, including older age, worse renal function at diagnosis, lower hemoglobin level and cardiovascular involvement, and a lack of non-glucocorticoid maintenance therapy, were independent predictors of mortality. The FFS at initial diagnosis appears to be a better tool than BVAS in predicting mortality, especially for early mortality at 1 year, with an AUC value of 0.874.

FUNDING STATEMENT

There was no funding for this work.

CONFLICT OF INTEREST

All authors reported no conflicts of interest in relation to this article.

ORCID

Joshua Ka Ho Yeung  https://orcid.org/0000-0002-5855-1089

Joyce Kit Yu Young  https://orcid.org/0000-0003-2955-3154

APPENDIX