Development of a Kidney Prognostic Score in a Japanese Cohort of Patients With Antineutrophil Cytoplasmic Autoantibody Vasculitis

Abstract

Introduction

Glomerulonephritis is frequent in antineutrophil cytoplasmic antibody (ANCA)-associated vasculitis (AAV) and crucial to disease outcomes. We conducted a detailed assessment of renal pathology in Japanese patients with AAV, and developed a new score that would predict renal outcome.

Methods

Two hundred twenty-one patients who were diagnosed with AAV and underwent a kidney biopsy were enrolled. Data on glomerular, tubular, interstitial, and vascular lesions from kidney biopsies were analyzed; the 3 established classification and prognostic scoring systems (Berden Classification, Mayo Clinic/RPS Chronicity Score [MCCS], and ANCA Renal Risk Score [ARRS]) were validated. Further, we developed a new prognostic score by including variables relevant for Japanese patients with ANCA-glomerulonephritis.

Results

Median follow-up was 60 months (interquartile range: 6–60). End-stage kidney disease (ESKD) risk prediction by the MCCS and the ARRS was confirmed. Moreover, our analysis identified 4 items with significant ESKD risk prediction capacity, namely percentage of cellular, fibrocellular, and fibrous crescents; and sclerotic glomeruli. Based on our findings, we created a score evaluating the percentage of these lesions to total glomeruli, the Percentage of ANCA Crescentic Score (PACS). The area under the receiver operating characteristic (ROC) curve evaluating PACS was 0.783. The PACS had a comparable performance as the ARRS in predicting ESKD. The optimal PACS cut-off for ESKD risk over 60 months was 43%. In addition, the percentage of cellular crescents and presence of interstitial inflammation were independent predictors of kidney function recovery.

Conclusion

We developed a new score predicting renal prognosis using histopathological data of Japanese patients with ANCA-glomerulonephritis. Studies are needed to validate our results in international cohorts.

Graphical abstract

AAV is a necrotizing vasculitis of small-to-medium sized vessels characterized by ANCA positivity and by virtue, lacking immune complex deposits (“pauci-immune”). AAV encompasses 3 clinical phenotypes, namely microscopic polyangiitis (MPA), granulomatosis with polyangiitis (GPA), and eosinophilic GPA.¹ MPA is associated with a high rate of myeloperoxidase (MPO)-ANCA positivity, GPA is associated with a high rate of proteinase 3-ANCA positivity, and eosinophilic GPA is often ANCA-negative. In particular, the kidneys and lungs are common sites of onset, and may be seriously affected in this disease.^2,3 Although improved outcomes have been achieved through newer therapeutic approaches, prediction of disease outcomes, such as death and ESKD, remains a challenging issue.^4,5

Glomerulonephritis is a form of organ involvement frequently observed in AAV and crucial to disease outcomes.⁶ The characteristic renal pathology of this condition is crescentic necrotizing glomerulonephritis.⁷ To date, several attempts have been made to predict disease outcomes based on renal pathology. In 2010, Berden et al.⁸ proposed a histopathologic classification, dividing glomerular lesions into 4 classes: focal, crescentic, mixed, and sclerotic. These classes were shown to correlate with renal prognosis. The most promising prognosis was observed in the focal class, followed by the crescentic, mixed, and sclerotic classes. This classification has been validated in various studies.9, 10, 11, 12 In 2017, Sethi et al. proposed the MCCS, with the assessment items glomerulosclerosis, interstitial fibrosis, tubular atrophy, and arteriosclerosis.¹³ The total score on each assessment item was demonstrated to correlate with renal prognosis.¹⁴ In 2018, Brix et al.¹⁵ proposed the ARRS. This score was reported to correlate with renal prognosis when the assessment items included normal glomeruli, severity of interstitial fibrosis or tubular atrophy, and estimated glomerular filtration rate (eGFR) at the time of diagnosis.¹⁶ In addition, fibrinoid necrosis and tubulointerstitial inflammation, which are acute lesions of glomerulonephritis, have been reported to be associated with renal prognosis.¹⁷ The glomerular classification proposed by Berden et al.,⁸ the MCCS, and the ARRS have been highly valued and reported to be useful for determining renal prognosis.⁶

The clinical features of AAV vary among races and regions.¹⁸ In particular, the prevalence of proteinase 3-ANCA-positive GPA is high in Europe and the United States, but MPO-ANCA-positive MPA is more common in Japan.^19,20 Studies on renal pathological findings have also been conducted in Japan to validate the glomerular classification proposed by Berden et al.⁸ and the ARRS, and have demonstrated the correlation between glomerular classification and renal prognosis.21, 22, 23, 24 However, these studies had a small sample size, and no large cohort studies have been reported to date. Moreover, no studies have comprehensively analyzed important renal pathological findings, such as glomerular, interstitial, tubular, and vascular lesions.

To address these issues, we conducted a detailed assessment of renal pathology in Japanese patients with AAV, validated previously reported scoring systems, and attempted to develop a new score that would predict renal prognosis.

Methods

The participants of this study were patients who visited the University of Yamanashi, Chiba University, Matsudo City General Hospital, Shimoshizu National Hospital, Tokyo Metropolitan Tama Medical Center, or Yamanashi Prefectural Central Hospital between 2000 and 2019. These patients were diagnosed with AAV and underwent a renal biopsy. We performed renal biopsies when patients had a urinalysis test suspicious of the presence of ANCA-glomerulonephritis, that is positivity for occult blood or protein. All clinical phenotypes, classified by the 2022 American College of Rheumatology/European League Against Rheumatism classification criteria for MPA, GPA, and eosinophilic GPA, were included.25, 26, 27 We excluded patients with other forms of glomerulonephritis (e.g., concomitant IgA nephropathy, diabetic kidney disease, or antiglomerular basement membrane disease). Data on patient characteristics and clinical courses were retrospectively extracted from electronic medical records for up to 60 months from the time of diagnosis. The extracted data included age, sex, diagnosis, type of ANCA, C-reactive protein levels, serum creatinine levels, eGFR, urinalysis results, somatic symptoms, comorbidities, and treatments.²⁸ Regarding renal pathology, 3 nephrologists knowledgeable in the discipline reassessed and discussed glomerular, tubular, interstitial, and vascular lesions at the same time.^24,29, 30, 31 Regarding renal pathological findings, we referred to the report by Berden et al. and The Oxford classification of IgA nephropathy. Glomerular lesions were as follows: cellular crescent: extracapillary cell proliferation of more than 2 cell layers with >50% of the lesion occupied by cells; fibrocellular crescent: an extracapillary lesion comprising cells and extracellular matrix, with <50% cells and <90% matrix; fibrous crescent: >10% of the circumference of Bowman's capsule covered by a lesion composed of ≥90% matrix; global glomerulosclerosis: >80% of the glomerulus sclerosed; and segmental sclerosis: <80% of the glomerulus sclerosed. In addition, the presence or absence of the following findings was also evaluated: fibrinoid necrosis, tuft adhesions to Bowman's capsule, Bowman's capsule rupture, mesangial cell proliferation, glomerular collapse, and periglomerular fibrosis. Regarding the glomerular basement membrane, the presence or absence of the following findings was evaluated: rupture, double-contour appearance, thickening, or spike formation. Tubular atrophy, tubular necrosis, interstitial fibrosis, and interstitial inflammation were evaluated semiquantitatively by calculating the percentage or severity grade as 0: absent (0%), 1: mild (≤25%), 2: moderate (26%–50%), 3: severe (>50%). The tubulitis was assessed semiquantitatively with cell count or severity grade as 0: absent (no mononuclear cells in tubules), 1: mild (foci with 1 to 4 cells/tubular cross section or 10 tubular cells), 2: moderate (foci with 5 to 10 cells/tubular cross section), 3: severe (foci with >10 cells/tubular cross section, or the presence of at least 2 areas of tubular basement membrane destruction accompanied by ≥26% interstitial inflammation and moderate tubulitis elsewhere in the biopsy).^32,33 Finally, vascular fibrinoid necrosis was evaluated.^8,34 A minimum of 5 glomeruli were required for assessment.¹⁵ The data of 268 participants were extracted. After the exclusion of 36 patients with insufficient data on patient characteristics, 4 patients with a questionable diagnosis, and 7 patients with less than 5 glomerular lesions, the final analysis population comprised 221 patients (Supplementary Figure S1). We performed a comprehensive analysis of the renal pathology and validated the previously reported renal pathological scoring systems.^8,13,15 Further, we developed a new renal prognostic score by performing multivariate analysis with patient characteristics. ESKD was defined as a persistent eGFR of <15 ml/min per 1.73 m².³⁵

Statistical Analyses

We performed statistical analyses using SPSS software, version 27.0 J (IBM Japan, Tokyo, Japan). Renal survival rate was calculated using the Kaplan–Meier method, and the rates between groups were compared using the log-rank test. To assess potential risk factors for ESKD, time-to-event analysis was performed using Cox proportional hazards model. The dependent variable was the presence or absence of ESKD during the observational period of 60 months. The independent variables included percentage of cellular crescents to total glomeruli, percentage of fibrocellular crescents, percentage of fibrous crescents, percentage of sclerotic glomeruli, percentage of normal glomeruli, semiquantitative score according to 4 grades of tubulitis, tubular atrophy, tubular necrosis, interstitial inflammation, interstitial fibrosis, presence or absence of vascular fibrinoid necrosis, initial eGFR, age, and positive or negative MPO-ANCA. We used a logistic regression model to assess factors predicting kidney function recovery. This was defined as a change in eGFR by ≥10 ml/min per 1.73 m² 6 months after treatment initiation from the eGFR at the time of diagnosis with reference to previous research.^36,37 Cut-off values were calculated by ROC analysis. Two-sided P-values <0.05 were considered statistically significant.

Ethics

This study was approved by the ethics committee of the University of Yamanashi (reference number: 1597). We collected patient information according to the Declaration of Helsinki and the Ethical Guidelines for Medical and Health Research Involving Human Participants in Japan.

Results

Patient Characteristics

In Table 1, baseline demographics and characteristics of the 221 included patients are highlighted. The mean age was 70 years and 114 (51.6%) of patients were women. MPA was diagnosed in 193 patients (87.3%), GPA in 22 (9.9%), and eosinophilic GPA in 6 (2.7%). MPO-ANCA positivity was observed in 196 patients (88.7%) and proteinase 3-ANCA positivity in 22 patients (9.9%). At the time of diagnosis, median serum creatinine level was 1.6 mg/dl and median eGFR was 31 ml/min per 1.73 m². Apart from ANCA-glomerulonephritis, lung involvement was reported in 114 patients (51.6%) and general symptoms, such as fever, in 176 (79.6%). Treatment for the induction of remission included glucocorticoids with a maximum dose of 40 mg/d (equivalent to prednisolone) in 213 patients (96.4%), methylprednisolone pulse therapy in 55 (24.9%), i.v. cyclophosphamide therapy in 78 (35.3%), rituximab in 36 (16.3%), and therapeutic plasma exchange in 11 (4.9%). Further, a combination of different immunosuppressants was used in 146 patients (66.1%) (see Table 1). Median follow-up time was 60 months (interquartile range: 6–60). ESKD occurred in 61 out of 221 patients (27.6%). Death occurred in 62 of 221 patients (28%).

Table 1.

Baseline demographics and characteristics of evaluated patients at the time of kidney biopsy

Clinical characteristics (N = 221)
Age, mean (SD), years	70 (11.9)
Sex, woman, n (%)	114 (51.6)
Diagnosis, n (%)
MPA	193 (87.3)
GPA	22 (9.9)
EGPA	6 (2.7)
Types of ANCA, n (%)
MPO	196 (88.7)
PR3	22 (9.9)
Serum CRP, median (IQR), mg/dl	6.1 (0.9-11.9)
Serum Cr, median (IQR), mg/dl	1.6 (0.84-11.9)
eGFR, median (IQR), ml/min per 1.73 m2	31 (15.9-60.2)
Urinalysis
Occult blood, 0, 1+, 2+, 3+, n (%)	8 (3.7), 29 (13.1), 64 (30), 118 (53.4)
Protein, 0, 1+, 2+, 3+, 4+, n (%)	9 (4.1), 94 (42.5), 79 (35.7), 30 (13.6), 7 (3.2)
UPCR, median (IQR), g/g Cr	1.13 (0.5-2.1)
NAG, median (IQR), IU/l	19.8 (11.1-30.4)
β2MG, median (IQR), µg/l	2038 (488-9328)
Clinical features, n (%)
General symptoms	176 (79.6)
Skin	49 (22.2)
Mucosal	12 (5.4)
Joint	38 (17.2)
Eye	21 (9.5)
ENT	36 (16.3)
Cardiac	21 (9.5)
Abdominal	9 (4.1)
Peripheral nerve	44 (19.9)
CNS	10 (4.5)
Kidney	216 (97.7)
Pulmonary	114 (51.6)
Interstitial pneumonia	93 (42.1)
Alveolar hemorrhage	21 (9.5)
Comorbidities, n (%)
Hypertension	147 (66.5)
Diabetes mellitus	105 (47.5)
Dyslipidemia	95 (43.0)
Treatment
Patients receiving PSL, median (IQR), n (%)	213 (96.4)
Maximum dosage of PSL, median (IQR), mg/d	40 (30-50)
Concomitant use of immunosuppressant, n (%)	146 (66.1)
Patients receiving mPSL pulse, n (%)	55 (24.9)
Patients receiving IVCY, n (%)	78 (35.3)
Patients receiving rituximab, n (%)	36 (16.3)
Patients receiving plasma exchange, n (%)	11 (4.9)

ANCA, antineutrophil cytoplasmic antibody; β2MG, beta-2 microglobulin; CNS, central nervous system; Cr, creatinine; CRP, C-reactive protein; eGFR, estimated glomerular filtration rate; EGPA, eosinophilic granulomatosis with polyangiitis; ENT, ear, nose and throat; GPA, granulomatosis with polyangiitis; IQR, interquartile range; IVCY, intravenous cyclophosphamide pulse therapy; MPA, microscopic polyangiitis; MPO, myeloperoxidase; mPSL, methylprednisolone; NAG, N-acetyl glucosamine; PR3, proteinase 3; PSL, prednisolone; UPCR, urine protein-to-creatinine ratio.

Renal Pathological Findings

The mean number of glomeruli observed per biopsy was 24. Among the glomerular lesions, global sclerosis was observed in 188 patients (85.1%). Regarding the prevalence of crescents, cellular crescents were observed in 141 patients (63.8%), fibrocellular crescents in 95 (50%), and fibrous crescents in 19 (8.6%). The prevalence of other glomerular lesions is shown in Table 2. In Table 3, we show the mean percentage of the main histopathologic parameters per patient. The mean percentage of normal glomeruli was 31.7%, that of sclerosis was 23.8%, and that of crescents was 20.4%. Compared to previous reports, our cohort had a higher percentage of normal glomeruli and lower frequency of sclerosis and crescents.^38,39 One patient exhibited spikes and thickening of the basement membrane and was diagnosed with concomitant membranous nephropathy.⁴⁰ In addition, comprehensive analysis of tubular lesions revealed that almost all patients had some type of tubular abnormality. Interstitial inflammation and fibrosis were observed in ≥80% of patients, and 44 patients (19.9%) had vascular fibrinoid necrosis.

Table 2.

The percentage of patients presenting with distinct histopathologic findings on kidney biopsy

Histopathologic findings (N = 221)
Number of glomeruli, mean (SD)	24 (13.8)
Glomerular lesions, n (%)
Sclerosis
Global sclerosis	188 (85.1%)
Segmental sclerosis	29 (13.1%)
Crescents
Cellular crescents	141 (63.8%)
Fibrocellular crescents	95 (50%)
Fibrous crescents	19 (8.6%)
Fibrinoid necrosis	77 (34.8%)
Tuft adhesions to Bowman's capsule	79 (35.7%)
Bowman's capsule rupture	7 (3.2%)
Mesangial cell proliferation	27 (12.2%)
Glomerular collapse	18 (8.1%)
Periglomerular fibrosis	18 (8.1%)
GBM
Rupture	21 (9.5%)
Double-contour appearance	6 (2.7%)
Thickening or spike formation	1 (0.4%)
Tubulointerstitial lesions, n (%)
Tubulitis, mild, moderate, severe	165 (74.7%), 130 (58.8%), 26 (11.8%), 9 (4.1%)
Tubular atrophy, mild, moderate, severe	213 (96.4%), 51 (23.1%), 63 (28.5%), 99 (44.8%)
Tubular necrosis, mild, moderate, severe	183 (82.8%), 73 (33.3%), 47 (21.3%), 63 (28.5%)
Interstitial inflammation, mild, moderate, severe	178 (80.5%), 54 (24.4%), 64 (28.9%), 60 (27.1%)
Interstitial fibrosis, mild, moderate, severe	203 (91.9%), 56 (25.3%), 52 (23.5%), 95 (42.9%)
Vascular fibrinoid necrosis	44 (19.9%)

GBM, glomerular basement membrane.

Tubular atrophy, tubular necrosis, interstitial fibrosis, and interstitial inflammation were evaluated semi-quantitatively. The following percentage or severity grading was used: 0: absent (0%), 1: mild (≤25%), 2: moderate (26%–50%), 3: severe (>50%). The tubulitis was assessed semi-quantitatively with cell count or severity grade as 0: absent (no mononuclear cells in tubules), 1: mild (foci with 1 to 4 cells/tubular cross section or 10 tubular cells), 2: moderate (foci with 5 to 10 cells/tubular cross section), 3: severe (foci with >10 cells/tubular cross section, or the presence of at least two areas of tubular basement membrane destruction accompanied by ≥26% interstitial inflammation and moderate tubulitis elsewhere in the biopsy).

Table 3.

Mean percentage of each glomerular lesion to the total number of glomeruli

Histopathologic findings (N = 221)	Mean	SD
Normal %	31.7	23.4
Total sclerosis %	23.8	22.4
Global sclerosis %	22.9	22.3
Segmental sclerosis %	0.9	2.9
Total crescents %	20.4	19.3
Cellular crescents %	11.2	13.6
Fibrocellular crescents %	5	8.3
Fibrous crescents %	0.5	1.9
Fibrinoid necrosis %	3.8	8.2
Tuft adhesions to Bowman's capsule %	3.9	7.0
Bowman's capsule rupture %	0.4	2.8
Mesangial cell proliferation %	1.4	5.2
Glomerular collapse %	0.6	2.5
Periglomerular fibrosis %	0.5	2.1
Glomerular basement membrane rupture %	1.5	7.3

Validation of the Previously Reported Glomerular Classifications

Patients were assessed based on the histopathologic classification proposed by Berden et al. Among the 221 patients, 32 patients were assigned to the sclerotic class (14.5%), 56 to the focal class (25.3%), 11 to the crescentic class (5%), and 122 to the mixed class (55.2%). In each glomerular class, the presence or absence of ESKD over a period of 60 months was analyzed using the Kaplan–Meier method. The focal and mixed classes were associated with a favorable prognosis, whereas the crescentic and sclerotic classes were associated with a poor prognosis (Figure 1a). Following this, we validated the MCCS. Minimal chronic changes were observed in 8 patients (3.6%), mild chronic changes in 80 (36.2%), moderate chronic changes in 90 (40.7%), and severe chronic changes in 43 (19.5%). As expected, minimal chronic changes were associated with the most favorable renal prognosis, whereas severe chronic changes were associated with the worst renal prognosis (Figure 1b). Severity according to the MCCS naturally reflected the risk of ESKD in this cohort. Finally, we validated the ARRS. Low risk was observed in 17 patients (7.7%), medium risk in 147 (66.5%), and high risk in 57 (25.8%). The risk stratification according to the ARRS predicted the risk of ESKD in this cohort (Figure 1c). In general, our analysis of the different scores in large Japanese population with ANCA-glomerulonephritis revealed a good applicability of currently used risk scores, whereas the limited sample of crescentic class patients in Japan might limit the generalizability of the Berden classification.

Figure 1.

Kaplan-Meier curves show renal survival according to previously reported histopathologic classifications in 221 Japanese renal biopsy patients with AAV. (a) Kaplan-Meier curves show renal survival according to Berden classification. (b) Kaplan-Meier curves show renal survival according to the Mayo Clinic chronicity score. (c) Kaplan-Meier curves show renal survival according to ANCA renal risk score. Renal survival is defined as end-stage kidney disease (ESKD) and/or eGFR<15 ml/min per 1.73 m². Renal survival rates were calculated using the Kaplan-Meier method, and the rates between groups were compared by the log-rank test. P values < 0.05 were considered statistically significant. AAV, ANCA-associated vasculitis; ANCA, antineutrophil cytoplasmic antibody

Development of a New Renal Prognostic Score

This study included a comprehensive analysis of renal pathological findings, and we also focused on all glomerular, tubular, interstitial, and vascular lesions, and their potential prediction of ESKD during an observational period of 60 months. The tubular, interstitial, and vascular lesions did not predict ESKD, whereas 4 remained independent predictors of ESKD after adjustment: percentage of cellular crescents (hazard ratio [HR], 1.029; P = 0.004; 95% confidence interval [CI]: 1.009–1.049), percentage of fibrocellular crescent (HR, 1.030; P = 0.033; 95% CI: 1.002–1.059), percentage of fibrous crescent (HR, 1.105 P = 0.031; 95% CI: 1.009–1.210), percentage of sclerosis (HR, 1.033; P < 0.001; 95% CI: 1.022–1.044) (Table 4). Clinical factors such as initial eGFR, age, and MPO were not independently predicting ESKD and were thus excluded. Based on these results, we attempted to develop a new renal prognostic score, which comprised the sum of the percentages of cellular crescents, fibrocellular crescents, fibrous crescents, and sclerotic glomeruli to total glomeruli. The risk score was termed the PACS. An ROC curve was plotted for outcomes of ESKD risk over 60 months. The area under the ROC curve was 0.783 (CI: 0.715–0.852) (Figure 2). The optimal cut-off PACS was 43%, with a sensitivity of 77% and a specificity of 68%. Particularly, the risk of ESKD over a period of 60 months increases when the PACS is 43% or higher. We focused on MPO-ANCA patients only, but the prediction of PACS remained the same.

Table 4.

Time to event analysis for end-stage kidney disease by Cox proportional hazards model in 221 Japanese patients with ANCA-glomerulonephritis

Variables	HR	95% CI		P value
Percentage of cellular crescent	1.029	1.009	1.049	0.004
Percentage of fibrocellular crescent	1.030	1.002	1.059	0.033
Percentage of fibrous crescent	1.105	1.009	1.210	0.031
Percentage of sclerosis	1.033	1.022	1.044	0.000

ANCA, anti-neutrophil cytoplasmatic antibody; ESKD, end-stage kidney disease; HR, hazard ratio; MPO, myeloperoxidase; 95% CI, 95% confidence interval.

The dependent variable was presence or absence of ESKD during the observational period of 60 months. The independent variables included percentage of cellular crescents to total glomeruli, percentage of fibrocellular crescents, percentage of fibrous crescents, percentage of sclerotic glomeruli, percentage of normal glomeruli, semi-quantitative score according to 4 grades of tubulitis, tubular atrophy, tubular necrosis, interstitial inflammation, interstitial fibrosis, presence or absence of vascular fibrinoid necrosis, initial eGFR, age, and positive or negative MPO-ANCA in each patient. Cox proportional hazards model was performed. P values < 0.05 were considered statistically significant.

Figure 2.

Receiver operator characteristic curves for the PACS and ESKD risk over 60 months in 221 Japanese renal biopsy patients with AAV. The percentage of ANCA crescentic score (PACS) is the sum of the percentages of cellular crescents, fibrocellular crescents, fibrous crescents, and sclerotic glomeruli to total glomeruli. AAV, ANCA-associated vasculitis; ANCA, antineutrophil cytoplasmic antibody; ESKD, end-stage kidney disease; PACS, percentage of ANCA crescentic score.

The counterpart of the glomerular item in the PACS is the normal glomerulus. The percentage of normal glomeruli was not identified as an independent significant item. We also compared the ROC curve for outcomes of ESKD risk over 60 months between the percentage of PACS and normal glomeruli. The area under the ROC curve was 0.783% (CI: 0.715–0.852) for PACS and 0.735% (CI: 0.662–0.808) for the percentage of normal glomeruli. This difference was not significant between PACS and the percentage of normal glomeruli (P = 0.09 (Figure 3). ESKD was predicted best by ARRS (area under the ROC curve 0.827; 95% CI: 0.765–0.889), second best by PACS (0.783; 95% CI: 0.715–0.852), and worst by MCCS (0.694; 95% CI: 0.619–0.768). However, there was no difference between PACS and ARRS (P = 0.126), whereas this was significant when comparing PACS and MCCS (P = 0.013) (Figure 4a–c). Therefore, the PACS might be similarly efficient to predict ESKD compared to the ARRS, by only focusing on histopathological changes and not incorporating clinical variables.

Figure 3.

Comparison of ROC curves of ESKD risk over 60 months between PACS and percentage of normal glomeruli in 221 Japanese renal biopsy patients with AAV. The area under the ROC curve was 0.783% (CI: 0.715–0.852) for PACS and 0.735% (CI: 0.662–0.808) for the percentage of normal glomeruli. There was no significant difference between PACS and the percentage of normal glomeruli (P = 0.09). The percentage of ANCA crescentic score (PACS) is the sum of the percentages of cellular crescents, fibrocellular crescents, fibrous crescents, and sclerotic glomeruli to total glomeruli. AAV, ANCA-associated vasculitis; ANCA, antineutrophil cytoplasmic antibody; CI, confidence interval; ESKD, end-stage kidney disease; PACS, percentage of ANCA crescentic score; ROC, receiver operator characteristic curves.

Figure 4.

Comparison of ROC curves of ESKD risk over 60 months between PACS and previously reported outcomes in 221 Japanese renal biopsy patients with AAV. (a) Comparison of ROC curves of ESKD risk over 60 months for PACS between PACS and ARRS. (b) Comparison of ROC curves of ESKD risk over 60 months for PACS between PACS and MCCS. (c) Comparison of ROC curves of ESKD risk over 60 months for PACS between ARRS and MCCS. AUC was the best for ARRS 0.827 (95% CI: 0.765–0.889), second for PACS 0.783% (CI: 0.715–0.852), and worst for MCCS 0.694 (95% CI: 0.619–0.768). There was no significant difference between PACS and ARRS (P = 0.126) (Figure 4a). There were significant differences between PACS and MCCS (P = 0.013) (Figure 4b), ARRS and MCCS (P = 0.018) (Figure 4c). The percentage of ANCA crescentic score (PACS) is the sum of the percentages of cellular crescents, fibrocellular crescents, fibrous crescents, and sclerotic glomeruli to total glomeruli. AAV, ANCA-associated vasculitis; ANCA, antineutrophil cytoplasmic antibody; ARRS, ANCA renal risk score; AUC, area under the ROC curve; CI, confidence interval; ESKD, end-stage kidney disease; MCCA, Mayo Clinic/RPS chronicity score; PACS, percentage of ANCA crescentic score; ROC, receiver operator characteristic curves.

We also analyzed the impact of different therapies on the predictive value of the PACS. The proportion of ESKD by treatment was lowest for the rituximab group 8.3% (3 of 36 cases), followed by the intravenous cyclophosphamide group 21.7% (17 of 78 cases) and the group not using both 38.8% (41 of 107 cases). The area under the ROC curve was 0.798% (CI: 0.398–1.197) for the rituximab group, 0.755% (CI: 0.634–0.875) for the intravenous cyclophosphamide group and 0.771% (CI: 0.678–0.863) for the group not using either therapeutic options (Supplementary Figure S2A–C).

Analysis of Factors Predicting Kidney Function Recovery

Serum creatinine levels and eGFR improved after the induction of remission treatment in some patients. The predefined change in eGFR 6 months after treatment initiation by ≥10 ml/min per 1.73 m² from the eGFR at the time of diagnosis was reached by 58 patients. The most significant predictor for kidney function recovery was the percentage of cellular crescents (odds ratio, 1.043; P = 0.003; 95% CI: 1.020–1.067), whereas borderline significance was observed for interstitial inflammation (odds ratio, 1.392; P = 0.045; 95% CI: 1.007–1.925) (Table 5).

Table 5.

Analysis of factors predicting kidney function recovery by logistic regression model in Japanese patients with ANCA-glomerulonephritis

Variables	Odds ratio	95% CI		P value
Percentage of cellular crescents	1.043	1.020	1.067	0.003
Interstitial inflammation	1.392	1.007	1.925	0.045

ANCA, antineutrophil cytoplasmatic antibody; eGFR, estimated glomerular filtration rate; MPO, myeloperoxidase; 95% CI, 95% confidence interval.

The dependent variable was change (Δ) in eGFR 6 months after treatment initiation by ≥10 ml/min per 1.73 m² from the eGFR at the time of diagnosis. The independent variables included percentage of cellular crescents to total glomeruli, percentage of fibrocellular crescents, percentage of fibrous crescents, percentage of sclerotic glomeruli, percentage of normal glomeruli, semiquantitative score according to 4 grades of tubulitis, tubular atrophy, tubular necrosis, interstitial inflammation, interstitial fibrosis, presence or absence of vascular fibrinoid necrosis, initial eGFR, age, and positive or negative MPO-ANCA in each patient. Logistic regression model was performed. P values < 0.05 were considered statistically significant.

Analysis of Risk Factors for Mortality

In a next step, we asked whether clinical characteristics and histopathological changes are capable of predicting mortality. No significant association between ESKD and death could be identified, which was probably impacted by the limited ESKD event rate in our study. Vascular fibrinoid necrosis emerged as the only histopathological risk factor predicting mortality; however, this result only yielded borderline significance (HR, 2.093; P = 0.036; 95% CI: 1.049–4.176) (Table 6).

Table 6.

Time to event analysis for mortality with Cox proportional hazards model in 221 Japanese renal biopsy patients with AAV

Variable	HR	95% CI		P value
Vascular fibrinoid necrosis	2.093	1.049	4.176	.036

AAV, ANCA-associated vasculitis; HR, hazard ratio; MPO, myeloperoxidase; 95% CI, 95% confidence interval.

The dependent variable was presence or absence of death over 60 months. The independent variables included percentage of cellular crescents to total glomeruli, percentage of fibrocellular crescents, percentage of fibrous crescents, percentage of sclerotic glomeruli, percentage of normal glomeruli, semi-quantitative score according to 4 grades of tubulitis, tubular atrophy, tubular necrosis, interstitial inflammation, interstitial fibrosis, presence or absence of vascular fibrinoid necrosis, initial eGFR, age, and positive or negative MPO-ANCA in each patient. Cox proportional hazards model was performed. P values < 0.05 were considered statistically significant.

Discussion

In this large-scale study involving Japanese patients with AAV, we assessed renal pathology in great detail and examined the associations between pathological findings and disease outcomes. The results showed that disease severity according to all previously reported classifications, which were the histopathologic classification proposed by Berden et al., the MCCS, and the ARRS, were generally consistent with renal prognosis. Following this, we developed a new renal prognostic score, the PACS, and examined risk factors for kidney function recovery.

Berden et al. indicated that renal prognosis worsened in the following order: focal, crescentic, mixed, and sclerotic classes. However, several subsequent reports and meta-analyses have shown that renal prognosis for the mixed and crescentic classes was similar.9, 10, 11, 12 Our analysis also revealed a better prognosis for the mixed class than for the crescentic class. Notably, only 5% of patients in the current study had a crescentic class and this might have influenced our results. This difference may also reflect that a higher percentage of Japanese patients are MPO-ANCA-positive. Chronic changes, such as glomerulosclerosis, interstitial fibrosis, and tubular atrophy are more commonly reported in MPO-ANCA-positive vasculitis than in proteinase 3-ANCA-positive vasculitis.^12,22,38,41, 42, 43 When patient characteristics were compared between our study and that of Berden et al.,⁸ our cohort included more elderly patients (62 years vs. 70 years, mean), and patients had a higher prevalence of hypertension (66.5%), diabetes (47.5%), and dyslipidemia (43%). Younger patients were also reported by Boudhabhay et al. Two hundred fifty-one patients, of whom 22.7% fell into the sclerotic class, 33.9% the focal class, 19.5% the crescentic class, and 23.9% the mixed class according to the glomerular classification, were included. The mean age of the patients was 63 years, and the prevalence of comorbidities was 37.8% for hypertension and 9.6% for diabetes.⁴⁴ These significant differences seem to have contributed to the difference in the proportion of mixed class patients, as observed in mainly European cohorts in comparison to our study. However, clinical characteristics such as age and MPO-ANCA were not predicting ESKD, thus were not incorporated into PACS. Similarly, common comorbidities such as diabetes, hypertension, and hyperlipidemia did not emerge as independent variables.

The 4 stages of the MCCS also correlated with renal prognosis in our cohort. Our study is the first to assess the MCCS in a Japanese cohort. Comparing patient characteristics between the cohort studied by Casal Moura et al. and our cohort, our cohort was again older (64 years vs. 70 years, mean) and included a higher percentage of patients with MPA (47% vs. 87%) and patients positive for MPO-ANCA (51% vs. 89%). The MCCS was demonstrated to be effective in predicting the prognosis of AAV in Asians, in whom MPA is common.¹⁴

The 3 grades of the ARRS also correlated with renal prognosis in our cohort. When patient characteristics were compared between the report by Brix et al. and our cohort, our cohort was older (66 years vs. 70 years, mean) and included a higher percentage of patients positive for MPO-ANCA (51% vs. 89%).¹⁵ Despite differences in patient characteristics, such as race and types of ANCA, the ARRS has been validated to better reflect renal prognosis compared with the glomerular classification proposed by Berden et al.^24,31

Based on general differences in the presentation of Asian patients with AAV in comparison to Caucasian populations, a main aim was to develop a histologic risk score of relevance for this population. PACS is the sum of the percentages of cellular crescents, fibrocellular crescents, fibrous crescents, and global sclerotic glomeruli to the total number of glomeruli. We demonstrated that a PACS ˃43% was associated with increased risk of ESKD. Our study included a comprehensive analysis of glomerular, tubular, interstitial, and vascular lesions that has not yet been conducted in such a large number of patients, and our study reveals important renal pathological lesions associated with ANCA-glomerulonephritis outcome. The addition of tubular, interstitial lesions, and vascular lesions did not contribute to refinement of the score. This is in line with observations leading to the establishment of the Berden classification, because the authors did not find additional value in adding tubulointerstitial lesions. By using the histopathological classification, van Daalen et al.⁹ reported no difference in renal prognosis between the crescentic and mixed classes in a meta-analysis.^10,45 The crescentic class is better phenotyped because only cellular crescents contribute to this class, whereas the mixed class is more heterogeneous and lacks clear pathological characteristics. A 3-class classification, based on findings after initial publication of the Berden classification, was considered more appropriate than a 4-class classification. In crescentic class patients, the eGFR is lower at the time of diagnosis, but improves after appropriate induction treatment, and eGFR may become comparable with that of mixed class patients. For this reason, van Daalen et al.⁹ stated that renal prognosis is better reflected by classifying crescents into cellular, fibrocellular, and fibrous only, rather than by combining the crescentic and mixed classes into one class. In 2019, Xu et al.⁴⁶ assessed the 4-class classification, reporting that renal prognosis worsened when the combined percentage of fibrous crescents and global glomerulosclerosis was 32.6% or greater, and that recovery of kidney function could not be expected when it exceeded 47.9%. These reports, in part confirmed by our analyses, underline the potential applicability of PACS, which may be a useful tool to predict ESKD risk in Asian populations with ANCA-glomerulonephritis. The ARRS predicted ESKD best in our Japanese cohort, which in general presents at older age and a high frequency of MPO-ANCA positivity (Figure 4). The PACS had a comparable performance as the ARRS in predicting ESKD. Moreover, the PACS as a histopathological score may allow the same prediction in slowly-progressing MPO-ANCA cases as the clinico-pathological model ARRS. Because age and kidney function did not improve prediction further in PACS, the active component of cellular crescents may have a greater impact. This highlights the need to adjust prediction models to the different disease phenotypes.

Regarding eGFR before and after treatment, we identified the percentage of cellular crescents and interstitial inflammation as factors predicting eGFR recovery during the first 6 months. Hauer et al. reported that improvement in eGFR and percentage of normal glomeruli at the time of diagnosis were associated with improved renal prognosis.^38,47 In ANCA-glomerulonephritis, glomerular basement membrane ruptures due to severe inflammation triggers the formation of cellular crescents. Crescent formation up to the fibrocellular level is considered an acute lesion and a potentially reversible change.^48,49 Brilland et al.⁵⁰ pointed out the association of lower percentage of tubular atrophy and interstitial fibrosis with improvement of eGFR after therapy. He et al.⁵¹ reported that ANCA-glomerulonephritis was more responsive to treatment in patients with interstitial inflammation, suggesting an association between interstitial inflammation and improvement of eGFR. Previous investigations examined renal pathological findings after treatment, and reported that the resolution of interstitial inflammation was associated with eGFR recovery. This indicates that interstitial inflammation is responsive to treatment and is reversible.^52,53 Wen et al.⁵⁴ encountered a patient with tubulointerstitial nephritis without glomerulonephritis who exhibited cellular crescents. Based on their case, they reported that, in the acute phase, interstitial inflammation occurs first, followed by formation of cellular crescents.⁵⁵ Our results suggest that interstitial inflammation and cellular crescents are early pathological findings of ANCA-glomerulonephritis. Simultaneously, we demonstrated that these conditions were reversible and could improve with treatment.

This study has some limitations. First, it is a retrospective cohort study, which might have led to various biases. Kidney biopsies were not performed in all subjects with AAV presenting to the recruiting hospitals; for example, some patients were already undergoing kidney replacement therapy and were considered not to benefit from immunosuppressive therapy. Exclusion of such patients may have contributed to a certain degree of bias. Moreover, information on kidney replacement therapy was missing in some cases, and this was therefore not included in final analyses. Second, the nature of treatments was not taken into consideration. Sample size in rare disease research is always an issue, but we aimed to collect all cases who presented to the participating hospitals over a 20-year period. During this period, the treatment strategy for AAV in Japan changed from glucocorticoid monotherapy to a combination therapy incorporating potent immunosuppressives such as cyclophosphamide or rituximab. It is obvious that these changes are likely associated with an improved prognosis.⁵⁶

In the current study, we developed a new score enabling prediction of renal prognosis by incorporation of a variety of renal pathological data of ANCA-glomerulonephritis, which were derived from a Japanese cohort. A novel aspect of PACS is the focus on histopathological damage and its potential predictive capacity of a slowly-sclerosing phenotype with no rapid decline in kidney function. We also identified risk factors for kidney function recovery, a relevant end point considering the impact of kidney dysfunction on morbidity and mortality. These results need to be confirmed in other Asian populations and might be validated in larger, international cohorts.

Disclosure

AK received consultancy fees from CSL Vifor, Otsuka, AstraZeneca, GSK, Walden Biosciences, Catalyst Biosciences, and Delta4. All other authors declared no conflicting interests.

Supplementary Material

Figure S1. Flowchart of patient selection for the current study.

Figure S2. ROC curves of ESKD risk over 60 months for PACS by each treatment. (A) ROC curve of ESKD risk over 60 months for PACS in 36 cases treated with rituximab. (B) ROC curve of ESKD risk over 60 months for PACS in 78 cases treated with intravenous cyclophosphamide. (C) ROC curve of ESKD risk over 60 months for PACS in 107 cases treated without rituximab or intravenous cyclophosphamide.

The proportion of ESKD by treatment was lowest for rituximab group 8.3% (3 of 36 cases), followed by intravenous cyclophosphamide group 21.7% (17 of 78 cases) and group not using both 38.8% (41 of 107 cases), though a retrospective analysis. The area under the ROC curve was 0.798% (CI: 0.398–1.197) for rituximab group (Supplementary Figure 2A), 0.755% (CI: 0.634–0.875) for intravenous cyclophosphamide group (Supplementary Figure 2B) and 0.771% (CI: 0.678–0.863) for group not using both (Supplementary Figure 2C). The percentage of ANCA crescentic score (PACS) is the sum of the percentages of cellular crescents, fibrocellular crescents, fibrous crescents, and sclerotic glomeruli to total glomeruli.

AAV, ANCA-associated vasculitis; CI, confidence interval; ESKD, end-stage kidney disease; PACS, percentage of ANCA crescentic score; ROC, receiver operator characteristic curves.