Introduction

Anti-neutrophil cytoplasmic antibody (ANCA)-associated vasculitis (AAV) is a rare, autoimmune disease that is characterized by a high positive rate of ANCAs and necrotic vasculitis of small vessels [1,2,3,4,5,6]. AAV includes granulomatosis with polyangiitis (GPA), microscopic polyangiitis (MPA), and eosinophilic granulomatosis with polyangiitis (EGPA). Although the incidence of AAV was reported to be as low as approximately 1.2 per 100,000 person-years in both Europe and Japan, it leads to high mortality (18–44%) and end-stage renal disease (23–30%) in affected patients [7,8,9,10,11].

Although various types of biopsies are performed on patients with AAV to confirm diagnosis and predict prognosis [12, 13], they have some adverse events. For example, it was reported that haemorrhages requiring transfusions occur in 0.4–1.5% of patients with renal biopsies and pneumothoraxes occur in 20.2% of patients with lung biopsies [14, 15]. They are performed in practice and recommended by the European League Against Rheumatism (EULAR) and European Renal Association-European Dialysis and Transplant Association (ERA-EDTA) [16, 17]; however, these recommendations are based only on studies that revealed the diagnostic sensitivities of each type of biopsy [18,19,20,21,22].

Although biopsies have decent diagnostic performance for AAV, whether biopsies improve the prognosis of AAV patients remains unknown. It is not acceptable to conduct harmful procedures to patients without rationale. Considering the situation, it is crucial to investigate the association between biopsies and prognosis as the first step in developing a rationale for the efficacy of biopsies in the management of AAV. The objective of this study was to elucidate the association between biopsies and prognosis utilizing the Japanese healthcare record database.

Materials and methods

Study design and setting

This retrospective cohort study used the RWD database, which is managed by the Health, Clinic and Education Information Evaluation Institute (HCEI; Kyoto, Japan) [23, 24]. The HCEI is a not-for-profit research service foundation in Japan that is supported by Real World Data Co., Ltd. (Kyoto, Japan) for data collection and standardization. This database consists of the data of 20 million patients from 190 hospitals in Japan. It includes demographic data, diagnoses with the International Statistical Classification of Diseases and Related Health Problems, 10th Revision Codes (ICD-10 code), prescriptions, laboratory results, and procedures of inpatients and outpatients. These data are managed by allocating unique identifiers for each individual. No linkage with other databases was performed for this study.

Data collection and definitions

We included inpatients who were 18 years and older and had codes associated with a diagnostic code representing AAV. Only data about first admissions associated with AAV were utilized for this study. We defined first admissions as the earliest admissions in the database. The data were available only after hospitals implemented electronic medical records, and the median date of admission in the database was July 2013 (interquartile range, July 2009–July 2016). The definition of “admissions associated with AAV” was an admission that had the ICD-10 code (M301, M310, M313, M317, M318, N007, N017, N019, N037, N057, J991) as at least one of the following three definitions: main disease, main trigger of admission, or the disease that was the most resource intensive during admission, which were recorded in the discharge abstract of each patient. These three definitions of ICD-10 codes were assigned by the physicians in charge. We excluded patients who had prior histories of dialysis. Patients who were prescribed prednisolone or immunosuppressive agents such as cyclophosphamide, rituximab, cyclosporine, tacrolimus, and azathioprine until 30 days before admissions were also excluded because they could already have been treated for AAV. We identified biopsies by procedural codes in administrative claims data. In Japan, although lung biopsies and otorhinolaryngological biopsies have some health care fee codes, there was no specific code for renal biopsies at the time when these data were collected. In practice, “percutaneous needle biopsy” and “open abdominal biopsy” are substituted as renal biopsies. Therefore, we extracted “percutaneous needle biopsy” and “open abdominal biopsy” as renal biopsies. We identified the procedural code “tissue sampling” from the ear, nose, oral cavity, pharynx and larynx as otorhinolaryngological biopsies. We also extracted the procedural codes for “tissue sampling” from skin, muscle, and peripheral nerve as other biopsies. The procedural codes for data extractions are shown in the Supplementary Materials.

This study was approved by the Institutional Review Board at Kyoto University and did not need individual consent because the data were anonymized.

Variables

The primary outcome was the composite outcome of death and RRT. Secondary outcomes were death alone, RRT alone [25], use of cyclophosphamide (CPA), use of rituximab (RTX), use of steroid pulse therapy, maximum dose of oral steroids per day, and use of transfusion during admission. We chose the composite outcome of death and RRT as the primary outcome because death and RRT are plausibly competing risks with each other for AAV patients who suffer from severe renal failure, and they would occur sufficiently often [10, 25, 26]. Outcomes were identified by ICD-10 codes, procedural codes, laboratory data, and the Anatomical Therapeutic Chemical code, which is a code for prescriptions.

Statistical analysis

Continuous variables are reported as the mean and standard deviation, and categorical variables are reported as numbers and percentages. First, we described the characteristics of patients who underwent biopsies and those who did not. The characteristics of patients with each type of biopsy were also described. We performed a chi-square test for binary data and a t-test for continuous data as univariable analyses. We also used multivariable linear regression analysis for continuous data and multivariable logistic regression analysis for categorical data. In multivariable analysis, we used age, sex, body mass index (BMI) [27], haemoglobin, serum albumin on admission, type of ANCA, anti-GBM antibody, Charlson comorbidity index (CCI), and disease severity, which is determined by the clinical grading system and indicates the severity of patients with rapidly progressive glomerulonephritis (RPGN), as confounding variables. These variables were indicated as potential confounding factors in previous studies [28,29,30]. The clinical grading system, which includes serum creatinine, C-reactive protein (CRP), and the presence of pulmonary regions, can predict the long-term prognosis of patients with RPGN [2]. We adopted this system because it precisely reflects the Japanese population and can be fulfilled with only laboratory data and the existence of lung involvement, which can be found in this database.

We dealt with treatments such as RTX, CPA, and the maximum dose of oral steroids as outcomes because we hypothesized that treatments would be changed by the results of biopsies. We did not utilize treatments as confounding variables because they are intermediate factors for other prognostic outcomes. We performed multiple imputation analysis with the mice package of R to fill in the missing BMI data and some laboratory data on admission by creating 10 imputed datasets. We reported P-values with a two-sided test and considered P < 0.05 to indicate statistical significance. All statistical analyses were conducted using R ver. 3.5.3.

Post hoc sensitivity analysis

The proportion of ANCA-positive patients was estimated as the baseline characteristics of the patients, and nearly half of the patients were found to be ANCA-negative. The reason why a subset of patients without laboratory confirmation was diagnosed with AAV was likely due to the Japanese-specific guideline [31, 32]. In that guideline, the combination of the following three features may allow physicians in Japan to make a probable diagnosis of AAV, even with ANCA-negative patients. However, since it is not possible to determine how the diagnosis of AAV was actually made from our database, we conducted a post hoc analysis to investigate the impact of the misclassification of AAV patients on the results. In this sensitivity analysis, we made two assumptions: (1) The diagnosis of the patients in the biopsy group was correct. (2) More outcomes occur in patients with AAV than in patients who are misclassified as having AAV because the prognosis of AAV is generally poor [7, 8, 10].

Results

We identified 676 admissions that were assigned codes associated with AAV from April 2005 to March 2019. Of them, 290 admissions were excluded for reasons such as age, number of admissions, RRT before admission, and immunosuppressant prescriptions (Fig. 1). Finally, 386 patients were included in this study. During the first admission, 154 (39.9%) inpatients underwent biopsies, and 232 (60.1%) inpatients did not. Eighty-seven patients underwent renal biopsies, 31 patients underwent lung biopsies, 10 patients underwent otorhinolaryngological biopsies, and 53 patients underwent other biopsies. Twenty-one patients underwent two of four types of biopsies, and 3 patients underwent three of them during the first admission with AAV. The baseline characteristics of each group are shown in Table 1. Younger age and male sex were more frequently observed in the biopsy group. The CCI was significantly different between the two groups; patients with CCI ≥ 5 were observed more frequently in the biopsy group. Other characteristics were not different between the two groups.

Fig. 1
figure 1

Algorithm for selection of AAV patients

Full size image
Table 1 Baseline characteristics of patients with biopsy and without biopsy
Full size table

We show the baseline characteristics of patients with each type of biopsy in Table 2. The serum creatinine level of patients with renal biopsy tended to be higher than that of patients with other types of biopsies. The CRP levels of patients with otorhinolaryngological biopsies were higher than those of patients with other types of biopsies. Interstitial pneumonia was observed more frequently among patients with lung biopsy.

Table 2 Baseline characteristics of patients with each type of biopsy
Full size table

In Table 3, we show crude and adjusted odds ratios (ORs) for outcomes in both groups. The composite outcomes of death and RRT were observed in 7 (4.5%) patients in the biopsy group and 25 (10.8%) patients in the no-biopsy group (OR 0.39 [95% CI 0.17, 0.94], P = 0.01). Two (1.3%) patients in the biopsy group and 10 (4.3%) in the no-biopsy group (OR 0.33 [95% CI 0.07, 1.53], P = 0.22) died during the first admission. RRT was observed in 6 (3.9%) patients in the biopsy group and 16 (6.9%) patients in the no-biopsy group (OR 0.55 [95% CI 0.21, 1.43], P = 0.28). The mean maximum dose of oral steroids was 32.2 mg in the biopsy group and 20.1 mg in the no-biopsy group (mean difference 10.8 mg [95% CI 7.9, 16.4], P < 0.01). There was no significant difference in transfusion between the groups (OR 1.23 [95% CI 0.76, 1.99], P = 0.39). In the multivariable analysis, the adjusted OR of the composite outcome of death and haemodialysis remained lower in the biopsy group than in the no-biopsy group (OR 0.31 [95% CI 0.12, 0.79], P = 0.01). The maximum dose of oral steroids also remained significantly higher in the biopsy group (mean difference 10.8 mg [95% CI 6.4, 15.2], P < 0.01).

Table 3 Crude odds ratio comparing outcomes between the no-biopsy group and the biopsy group
Full size table

A post hoc sensitivity analysis to examine the impact of misclassification showed that the results did not differ depending on the presence of misclassification. The details of this analysis are shown in the Supplementary Material.

Discussion

In this study, we found that biopsies in inpatients with AAV were associated with composite outcomes of death and RRT. Notably, composite outcomes were observed more frequently in the no-biopsy group. This result was consistent between the univariable analysis and multivariable analysis. Additionally, the maximum dose of oral steroids was significantly higher in the biopsy group than in the no-biopsy group. As one of the surrogate outcomes of haemorrhage, transfusion was not significantly different between the biopsy group and the no-biopsy group. Our results showed that AAV patients with biopsies received more intensive treatments and had a better prognosis during admission.

To our knowledge, this is the first study to compare the short-term prognosis of AAV patients who underwent biopsies with those who did not. Previous studies focused only on patients who underwent biopsies and whose histopathologic patterns and prognosis were assessed [26, 27]. Although previous studies revealed the diagnostic performance of biopsies for AAV patients, there were no published data to indicate the association between biopsies and the prognosis of AAV patients. Moreover, in our study, we utilized laboratory data from the database to adjust the difference between the biopsy group and the no-biopsy group. Our study offered new findings in the management of AAV patients, which implies the benefit of performing biopsies.

The finding that the frequency of transfusion was not significantly different between the two groups was also noteworthy. Although the risk of haemorrhage as an adverse effect of biopsy discourages physicians from performing biopsies, our results may support performing biopsies in terms of safety. However, we must admit that the sample size may not be large enough to detect a difference in such a rare outcome between the two groups.

There are five limitations in this study. First, the present study was a retrospective study. There would be confounding by indications between the biopsy group and the no-biopsy group, and confounding would widen the gap between the two groups. However, we utilized the laboratory data and adjusted for it as much as possible in multivariable analysis. Second, there was no specific health care fee code for renal biopsy during the target period of this study. We substituted “percutaneous biopsy” and “open abdominal biopsy”, which could include other types of biopsies. However, given that the biopsies were performed during the first admission with AAV, we assume that most of them would be renal biopsies. Moreover, the fact that patients with renal biopsies had higher serum creatinine levels indicated that the substitution of “percutaneous biopsies” and “open abdominal biopsy” for renal biopsy was valid (Table 2). Third, this study did not include admissions before hospitals applied the electronic medical record system. Second and subsequent admissions could have been identified as the first admission just after hospitals implemented electronic medical records. Therefore, we excluded patients who were treated with steroids or other immunosuppressive agents before the first admission with AAV. Thus, there may be some misclassification between the biopsy group and the no-biopsy group. However, we considered this type of misclassification to be nondifferential misclassification in this study because there is no relationship between the timing of admissions with AAV and the timing at which hospitals adopted electronic medical records. Among the situations with nondifferential misclassification, the direction of the bias would close the gap between the two groups [31]. Even in the situation, some outcomes were significantly different between the two groups. We believe the gap would remain significant and wider, even without misclassification, and the results with statistical significance are acceptable. Fourth, ICD-10 codes associated with AAV were not validated in Japan and thus may not completely reflect the actual disease. Regarding this limitation, we believe almost all of the codes reflected the actual disease because they were given by the doctors in charge in Japan. Finally, this study was based on the Japanese healthcare record database. These results may not be applicable to other regions.

In conclusion, we elucidated that AAV patients who underwent biopsies had a better prognosis in terms of the composite outcome of death and RRT in a Japanese healthcare database.