Increasing incidence and improved survival in ANCA-associated vasculitis—a Danish nationwide study

Abstract

Background

Anti-neutrophil cytoplasmic antibody (ANCA)-associated vasculitis (AAV) carries a high risk of morbidity and mortality, with outcomes modified by treatment and an incidence that may be increasing. We examined temporal changes in incidence and mortality during 2000–15 using nationwide healthcare registries.

Methods

Patients with incident AAV were identified using International Classification of Diseases Version 10 (ICD10) codes and grouped according to inclusion year (Period 1: 2000–04, Period 2: 2005–09, Period 3: 2010–15). Log link cumulative incidence regression adjusted for age, sex, renal function, cardiovascular disease, diabetes, hypertension and advanced disease severity were used to model survival.

Results

We identified 1631 patients (52% male), corresponding to an incidence of 18.5 persons/million/year (Period 1: 15.1, Period 2: 18.5, Period 3: 21.4). The slope of incident serologic ANCA testing was steeper than that of AAV (P = 0.002). Mean [standard deviation (SD)] age was 60.2 (16.7) years and mean (SD) follow-up was 6.8 (4.7) years. A total of 571 (35%) patients died (5-year mortality of 22.1%), with an absolute risk ratio (ARR) for Periods 2 and 3 compared with Period 1 of 0.80 [confidence interval (CI) 0.65–0.98, P = 0.031] and 0.39 (CI 0.31–0.50, P < 0.001). About 274 patients developed end-stage renal disease (ESRD) [16.8% (Period 1: 23.3%, Period 2: 17.6%, Period 3: 12.5%)], with ARR decreasing over time: Period 2 0.61 (CI 0.42–0.87, P = 0.007) and Period 3 0.57 (CI 0.39–0.83, P = 0.003). The overall risk of death associated with ESRD or chronic kidney disease was 1.74 (CI 1.29–2.37, P < 0.001) and 1.58 (CI 1.21–2.07, P < 0.001).

Conclusions

Incidence of ANCA testing and AAV diagnosis increased over the test period. Falls over time in mortality and ESRD risk may relate to earlier diagnosis and changes in treatment practice.

INTRODUCTION

Granulomatosis with polyangiitis (GPA) and microscopic polyangiitis (MPA) are the major subgroups of anti-neutrophil cytoplasmic antibody (ANCA)-associated vasculitis (AAV) [1]. AAV is an autoimmune disease characterized by inflammation and necrosis of the small blood vessel wall with preferential involvement of the skin, respiratory tract and kidneys, but can affect any organ and carries a high risk of morbidity and mortality [2–4]. GPA and MPA display similar renal histopathological lesions, a pauci-immune necrotizing glomerulonephritis [5], share a putative ANCA-associated pathogenesis [6] and respond equally to contemporary induction treatment regimens. The two serologic subgroups of ANCA, proteinase 3 (PR3) and myeloperoxidase (MPO), have different genetic predisposition [7], differ in the risk of relapse [8] and the long-term risks of end-stage renal disease (ESRD) [9], mortality and cardiovascular disease (CVD) [10, 11].

Previous epidemiological studies have been challenged by the rare occurrence of AAV, and large nationwide populations of unselected AAV patients are not readily accessible. While there appears to be a latitudinal gradient with GPA more frequent in the North of Europe as well as a longitudinal gradient with MPA more frequent in Japan [12, 13], reports on the incidence of GPA to MPA in Europe have been inconsistent [14, 15], with a mismatch between stable AAV incidence and mortality rates [3, 14, 16, 17], but an increasing AAV prevalence [17]. Apart from sample size, follow-up time and case ascertainment, which can affect epidemiological results [18], regional variation in disease epidemiology may also bias the results, such as, local awareness and referral practice. In this regard, a recent report from the North of Norway highlighted an increased incidence in AAV over time [19].

Diagnostic and therapeutic advances implemented during the last decades have changed the paradigm of AAV from a fatal diagnosis to a chronic relapsing disease [3, 11, 20]. However, despite the positive impact on survival, contemporary treatment regimens are still associated with frequent adverse events, e.g. infection, malignancy and infertility; and the primary causes of death after initial diagnosis are associated with the immunosuppressive treatment rather than the disease itself [3, 21]. Incidentally, contemporary long-term outcome studies, evaluating incidence, treatment outcomes, morbidity and mortality are needed, as highlighted in the 2016 European League Against Rheumatism (EULAR)/ERA-EDTA recommendations [1]. In this study, we examined possible temporal changes in incidence and mortality of AAV during 2000–15 using Danish nationwide healthcare registries.

MATERIALS AND METHODS

Patients and registries

This retrospective cohort study included all patients diagnosed with at least two consecutive hospital encounters registered as GPA or MPA [International Classification of Diseases Version 10 (ICD10): DM31.3 and DM31.7] between 2000 and 2015, regardless of primary organ manifestation or severity, using the Danish nationwide administrative registries as data source. Patients were grouped according to year of inclusion (Period 1: 2000–04, Period 2: 2005–09, Period 3: 2010–15). The ICD10 diagnostic codes associated with necrotizing vasculitis (DM31) and the codes specific for GPA (DM31.3) and MPA (DM31.7) were assessed in a pilot study of 104 patients from three distinct centres in the eastern regions of Denmark and found to have a positive predictive value (PPV) of 87 and 98% in finding patients with necrotizing vasculitis, and AAV as a group, respectively (Supplementary Material). One hundred and four patients were identified via ICD10 codes in the National Patient Registry (DM31: 104, DM31.3 and DM31.7: 44) and manually cross-referenced in the patient charts. A chart diagnosis of AAV was considered true if there was a sensible disease presentation supported by one of the three criteria: (i) biopsy report; (ii) ANCA positivity and treated as AAV; and (iii) if the diagnosis was undertaken by specialty department and treated as AAV (approved by the Danish Data Protection Agency, no. RH-2017-306, I-Suite no. 05916). The ICD10 codes pertaining to eosinophilic GPA were not validated as part of the pilot study and hence, not included in the main study.

In Denmark, every citizen has a unique and permanent civil registration number that enables individual-level linkage of information from nationwide administrative registries. The National Patient Registry holds information on all admissions to Danish hospitals since 1978 according to ICD codes and Nordic classification of surgical procedures, respectively. Information on all prescriptions dispensed from Danish pharmacies since 1995 is captured in the Danish Registry of Medicinal Product Statistics (prescription registry) according to the international Anatomical Therapeutic Chemical classification system [22]. Information on medication given during hospital admissions is currently not accessible. Primary and contributing causes of death are registered in the National Causes of Death Registry, and information on birth date, vital status and gender were obtained from the Central Person Registry. Information on ANCA testing, including both indirect immunofluorescence and ELISA assays, was extracted from the laboratory database of the Northern Region of Denmark, ∼10% of the Danish general population from 2000 to 2012 [23]. To increase the sensitivity of the diagnostic codes, prescriptions of glucose-lowering drugs (A10) and loop diuretics (C03C) were used as proxies for diabetes and congestive heart failure (CHF) [24]. Similarly, we defined hypertension by treatment with two or more antihypertensive drugs within a period of 3 months, or any hospital admission with a hypertension diagnosis (ICD8: 400–404, ICD10: DI10–15) [25]. Baseline renal function was assessed by the need for dialysis (ESRD) or development of chronic kidney disease (CKD) within 30 days after initial hospital discharge, compared with no renal involvement (for the survival analyses, baseline renal function was evaluated at index date to avoid immortality bias).

Statistics

Crude incidence of AAV was expressed as cases/million/year. The chi-square test and Student’s t-test were used to examine differences in categorical and continuous variables. Pearson’s product-moment correlation coefficient was used to evaluate if two variables were correlated. Cumulative incidence was assessed by Gray’s non-parametric test as well as the Aalen–Johansen cumulative incidence estimator. Absolute risk regression models, based on log link cumulative incidence regression [26], were used to model survival expressed as absolute risk ratios (ARRs). In analyses of survival data, patients entered the model at the date of discharge from the time of first hospitalization when the AAV diagnosis was initially confirmed, and non-hospitalized patients diagnosed from outpatient clinics entered the model at the day of diagnosis. Patients were subsequently followed until a primary endpoint occurred (death or ESRD), a maximum of 5 years, or until 31 December 2017, whichever came first. The main analyses were adjusted for time, age, sex, length of initial hospital stay, renal involvement, CVD, hypertension and diabetes. Patients who died during initial hospitalization were excluded from the survival analysis, as well as information on intensive care unit (ICU) stay, which was not accessible for the full study period. Also, as plasma exchange and lung haemorrhage were insufficiently coded before 2004 in the Danish registries, these variables were likewise omitted in the statistical models. By use of the same registries as applied for the study cohort and Exposure Density Sampling as matching method, a background population matched 1:1 on age and sex was identified and included as control group in the assessment of mortality and risk of ESRD. Each matched control was given the same exposure date as their respective case. Analysis of ANCA testing in the Northern Region of Denmark from 2000 to 2012, encompassing an average population of 579 200, were used to mirror a national tendency of ANCA testing. A two-sided P $≤$0.05 was considered significant. Analyses and data management were performed in SAS version 9.4 (SAS Institute Inc., Cary, NC, USA) and R version 3.6.1 (R Foundation for Statistical Computing, Vienna, Austria).

Post hoc validation analysis

The primary study population and key outcomes (i.e. included patients, incidence and risk of death and ESRD) were tested in a validation cohort identified by the use of a new set of inclusion criteria, previously validated by Sreih et al. [27]. The validation cohort was identified in the National Patient Registry using a combination of (i) ICD10 codes (ICD10: DM31.3 and DM31.7), (ii) encounter type (either ‘inpatient’ defined as an ICD10 code submitted on three consecutive days or ‘outpatient’ defined as two ICD9 codes submitted 3 months apart) and (iii) specialty involved in the care (Rheumatology, Immunology, Nephrology, Pulmonology or Otorhinolaryngology). The latter criterion was slightly modified in this study as some Danish hospitals did not differentiate between the specialties within Internal Medicine. Accordingly, we included Internal Medicine in addition to the above mentioned specialties. Due to the risk of immortal time bias as consequence to the inclusion method, a 90-day quarantine period was applied (index + 90 days) in analyses of survival and risk of ESRD. Finally, in assessment of our primary study population, we examined the percentage of patients included from the northern part of Denmark that at least had one ANCA test (ELISA or immunofluorescence) done in association with their initial diagnosis.

Ethics

Retrospective registry-based studies including anonymized data do not require ethical approval in Denmark. This study was approved by the Danish Data Protection Agency (ref. GEH-2014-018, I-Suite no. 02736 and VD-2018-292, I-Suite no. 6536).

RESULTS

Population demographics—total population

Between 2000 and 2015, 1631 patients with incident AAV were identified from the Danish National Patient Registry corresponding to an overall incidence of 18.5 [standard deviation (SD) 3.6] cases/million/year and increasing over time (Figure 1). Frequency increased with age, with incidence for the total study period of 6.0 (SD 1.9), 36.6 (SD 5.3) and 51.0 (SD 21.2) for patients <45, 45–75 and >75 years, respectively. Moreover, the highest increase in incidence over time was seen among patients >75 years old (Figure 2). About 52% were male, mean (SD) age at the time of diagnosis was 60.2 (16.7) years, and mean (SD) follow-up was 6.8 (4.7) years (Table 1). About 333 (20.4%) patients were registered as having had an ICU stay, 425 (26%) patients met the criteria of advanced disease severity judged by the length of their initial hospital stay and 204 (12.5%) had commenced on haemodialysis (HD) at Day 30 after discharge. A total of 571 (35.0%) patients died during total study period corresponding to a 5-year mortality of 22.3%, and primary contributing causes of death were CVD (28.6%) cancer (16.7%), infection (15.1%) and airway disease (10.8%). Cumulative frequencies of AAV diagnoses did display some seasonal changes, with the lowest number of cases during Summer and the highest during Fall (Figure 3). Regional distribution of mean AAV incidences in Denmark is shown in Figure 4.

Population demographics—stratified in groups

There was an increase in incidence of AAV throughout the three time periods [Period 1: 15.1 (SD 2.8), Period 2: 18.6 (SD 2.2), Period 3: 21.4 (SD 2.9), P = 0.006]; these findings were sustained in all sub-analyses of gender (data not shown) and age groups (Figure 2). Furthermore, we found a >3-fold increase in incident ANCA tests (per million per year) in the northern administrative region of Denmark during 2000–12 [2000–04: 1098 (SD 239), 2005–09: 1516 (SD 208), 2010–12: 2030 (SD 278)]. The slope of the average increase in ANCA tests was greater than that of incident AAV diagnoses (P = 0.002), and the two variables were significantly correlated [correlation coefficient: 0.76 [95% confidence interval (95% CI) 0.37–0.93, P = 0.002] (Figure 1). Diagnoses of CKD at the time of inclusion increased while the proportion of patients commencing HD within 30 days after discharge and patients with a prolonged hospital stay lasting >10 days, decreased over time. Individuals identified with diabetes and hypertension also increased throughout the three time periods. The overall distribution of age and gender at the time of diagnosis remained statistically unchanged throughout the three time periods, although the share of patient >75 years increased significantly and there was a trend towards a greater percentage of women (Table 1). There was no difference regarding peripheral vascular disease (PVD), ischaemic heart disease (IHD), chronic obstructive pulmonary disease (COPD) or CHF (Table 1). The primary causes of death pertaining to the three periods are listed in Table 2.

Survival analyses

Of the 1631 patients with incident AAV, 23 were excluded due to either in-hospital death or coding errors [Period 1: 6 (1.48%), Period 2: 6 (1.18%), Period 3: 11 (1.53%), P-value for difference = 0.871], leaving 1608 patients for inclusion in the survival analysis. The overall risk of death associated with AAV was significantly increased as compared with the matched background population (ARR = 2.38, 95% CI 1.98–2.86, P < 0.001). Crude incidence rates of all-cause death expressed as deaths per 1000 person-years were 0.17 (0.14–0.21), 0.14 (0.12–0.17) and 0.12 (0.10–0.15) for Periods 1, 2 and 3, respectively, and in the adjusted main analysis, risk of death following an AAV diagnosis decreased over time with an ARR for Periods 2 and 3 as compared with Period 1 of 0.79 (95% CI 0.65–0.97, P = 0.022) and 0.58 (95% CI 0.47–0.0, P < 0.001) (Figure 5A). Of the other covariates included in the model, higher age (ARR = 1.06, 95% CI 1.05–1.07, P < 0.001), CVD (ARR = 1.50, 95% CI 1.19–1.88, P < 0.001) and impaired renal function, were significantly associated with increased mortality, where ARR for death associated with the need of dialysis, and development of CKD at the day of initial hospital discharge compared with no renal involvement was 1.74 (95% CI 1.42–2.12, P < 0.001) and 1.37 (95% CI 1.13–1.66, P = 0.001), respectively. Length of initial hospital stay (ARR = 1.06, 95% CI 0.83–1.35, P = 0.636), male sex (ARR = 1.20, 95% CI 1.03–1.41, P = 0.0.023), hypertension (ARR = 1.18, 95% CI 1.00–1.40, P = 0.049) and diabetes (ARR = 1.12, 95% CI 0.78–1.62, P = 0.544) did not predict death in the current model. Also, the frequency of ESRD decreased during the three time periods (Period 1: 23.3%, Period 2: 17.6%, Period 3: 12.5%), corresponding to an ARR for Period 2 of 0.61 (95% CI 0.42–0.87, P = 0.007) and Period 3 of 0.57 (95% CI 0.39–0.83, P = 0.003) as compared with Period 1 (Figure 5C). Finally, there was a significantly increased risk of death associated with all strata of renal involvement when including the general population matched on age and sex as reference (HD at discharge: ARR = 3.52, 95% CI 2.83–4.37, P < 0.001; CKD at discharge: ARR = 2.58, 95% CI 2.07–3.22, P < 0.01; no renal involvement at discharge: ARR = 1.98, 95% CI 1.60–2.45, P < 0.001) (Figure 5B).

Validation analyses

By applying two of the three validated inclusion criteria (ICD10 codes and encounter type), we identified 1518 patients with AAV corresponding to 93.1% of the original study population; by additionally including the third criterion, i.e. specialty involved in the care, 1481 patients were identified as having AAV, corresponding to 90.8% of the original study population (all patients identified were contained in the original cohort). A total of 57 patients died during the inclusion period and hence, did not meet the validation inclusion criteria. The overall AAV incidence based on the three validated inclusion criteria was 16.9 (SD 2.9), with a significant increase over time [P = 0.023 (Supplementary data, Figure 2)], and risk of death and ESRD were decreasing over time (Supplementary data, Table S3). Finally, 93% of the original study patients identified in the Northern Region of Denmark, had at least one ANCA test done in association with their diagnosis.

DISCUSSION

We investigated whether the incidence and outcomes of AAV were time dependent using nationwide registries, and whether this may have been influenced by the frequency of ANCA testing. We found an increasing incidence of AAV between 2000 and 2015 in the total population and in subgroups stratified on age and gender; incidence was markedly higher for those aged >75 years. Some national regional and periodical variations were observed in AAV incidences as well as seasonal variation with highest and lowest cumulative frequencies of incident AAV diagnoses during Fall and Summer, respectively. The increasing incidence of AAV was associated with a marked increase in frequency of ANCA testing. Frequencies of major causes of death, i.e. CVD, infection, malignancy and respiratory disease, had an overall decreasing trend. Cumulative incidence of death, as well as risk of death and ESRD, declined throughout the three time periods, and main predictors of death, other than the year of inclusion, were higher age, CVD and renal involvement. The inclusion method defining our primary study population as well as key outcomes were successfully authenticated in a validation cohort.

It is generally recognized that incidence of AAV up until 2000 has been increasing [28, 29], putatively due to increased awareness, and specifically due the introduction of immunofluorescence assays for ANCA-testing during the 1980s and later the more specific anti-MPO and anti-PR3 ELISA assays during the 1990s [30]. More recent reports have claimed that incidence had plateaued around 15–20/million/year as the effect from the introduction of ANCA testing has ceased [17, 31]. In this study, supported by a recent Norwegian publication [19], we report a consistent increase in AAV incidence between 2000 and 2015, peaking in 2015 with an incidence of ∼25/million/year. Additionally, this coincided an even steeper slope in incident ANCA testing in the same period, suggesting that the effect from ANCA testing is still relevant. In this regard, the commercialization of the anti-MPO and anti-PR3 assays with the introduction of third-generation ELISA kits during the early 2000s [30, 32] contributed to widespread availability in local laboratories, and it is also likely that changes in physician education and training will have been reflected in assay requests. Furthermore, increasing incidence was consistent across all age and gender strata tested, with a marked augmented slope velocity among patients of older age. This finding further supports the notion that increasing ANCA testing has influenced the incidence, as the likelihood of identifying incident AAV episodes is greater in a population where the a priori risk of the disease is high [14, 15, 20].

There was some degree of seasonal variation in the frequency of new AAV diagnoses during the study period; however, this finding merely represents a description of the figures over time and does not pose any causal explanations. In this regard, although we cannot exclude some form of causal link between season and AAV pathogenesis, a low incident AAV frequency during Summer with a compensational higher frequency during Fall could be explained by social behaviour and reduced elective healthcare services during the former period. As also noted by others [19], there appeared to be a time-dependent variation in AAV incidence over time with an alternating zenith and nadir every 3–4 years (Figure 1). Furthermore, incidence varied across the 98 Danish administrative regions analysed (Figure 4) putatively due to changes in the demographic composition throughout the country. Temporal and regional variations in incidence may pose an epidemiological challenge for studies restricted to a single region in a limited period of time, and may explain the previous reported stable incidence around 15–20/million/year [17, 31] rather than an increasing incidence within this range suggested in this study. This may also apply to a previous American study that found stable, however relatively high, incidences over a 20-year period (1996–2015), as data were restricted to one region only [31]. Time-dependent changes in the demographic composition may likewise translate into altered disease frequencies, especially in the setting of the age-dependent increase in AAV incidence implied by our findings and others [14, 15, 20]. Accordingly, as the demography is changing towards higher percentages of people >75 years of age in Denmark [33], we expect future incidence to rise, disregarding the effect from serologic testing, which should fade over time.

There has been an increasing interest in AAV during recent decades. Research activity has increased and dedicated international research collaborations have facilitated development of disease definitions and classification criteria [34], guidelines [1] and treatment regimens, with specific focus on reduction in cumulative doses of cyclophosphamide [20, 35] and steroids [36], and introduction of new steroid-sparing agents, i.e. anti-CD20 targeted therapy with monoclonal antibodies [37]. However, although mortality rates have improved with the introduction of steroids and cyclophosphamide [38], and more recent epidemiological studies have suggested improved survival and renal outcomes [20, 39, 40] as a consequence of improved awareness, earlier disease detection and better treatment, we are the first to present head-to-head comparison of unselected nationwide epidemiological data from three consecutive time periods between 2000 and 2015 to support these findings, with a clear decrease in risk of ESRD and death following incident AAV diagnoses. The overall 5-year mortality rate of 22.3% was in keeping with previous studies [3, 16], and expectedly better than studies done in populations that exclusively included patients with renal involvement [39, 40]. However, our results did not support equal risks of mortality between AAV patients without real involvement and the general population as recently suggested [41].

Concordant with other studies, we found higher age, the presence of CVD and renal involvement to predict death [3, 16, 42]. Hypertension and diabetes did not predict mortality, possibly as the risk from these two variables is overshadowed by that of AAV itself, as well as treatment-related side effects [21]. In this regard, the autoimmune inflammatory response may be a major risk factor in developing CVD as noted with other autoimmune disease entities, e.g. rheumatoid arthritis, psoriasis and inflammatory bowel disease [43, 44], which may explain why CVD remains a high-risk factor in patients with AAV [3, 16]. Accordingly, previous studies have found significantly increased risk of CVD in patients with AAV and that this risk not only was associated with the traditional risk factors but also disease-related factors defined by, e.g. a higher Birmingham Vasculitis Activity Score [45–47]. Our findings may have been mediated through better baseline comorbidity status, including better renal function, simply due to earlier disease detection. However, as the main statistical model accounted for possible inequalities in baseline parameters with potential impact on main outcomes, our findings are most likely attributed to other features as well, e.g. more streamlined and effective AAV treatment regimens [1], which would be in keeping with the increasing prevalence of AAV as reported previously [17]. In this regard, we found a decreasing trend in death caused by CVD, cancer, infection and pulmonary disease, which previously have been noted as the primary causes of death in this population [3, 16]. In general, better treatment and awareness of such comorbidities with high impact on mortality in patients with AAV may have added to better survival [48, 49], although a more detailed assessment of mortality causes and temporal trends is warranted, especially to evaluate long-term outcomes from anti-CD20 target therapies as well as treatment strategies that are yet to be implemented, e.g. complement targeted therapies [50].

Limitations

Despite adjustment for some but not all factors are known to influence the risk of ESRD and death after an AAV diagnosis, residual confounding is inherent to observational studies and epidemiological associations do not per se represent causal associations. Furthermore, results might have been affected by limited sensitivity of the diagnostic codes by which AAV diagnoses, risk factors and causes of death were defined, and confounding may have been introduced by the unequal distribution of baseline parameters, although most of these were accounted for in the statistical model. The diagnostic codes used in this study (DM31.3 and DM31.7) yielded a PPV of 98% when used to find patients with AAV as a group. However, the codes did not perform well in differentiating between GPA and MPA, as most cases were coded as DM31.3. Also, we did not have access to histological, biochemical or serological data. Hence, interpretation of the study results may have been blurred by the inability to discriminate between specific diagnoses (GPA, MPA), serology (PR3, MPO) and eGFR, although we did include surrogate markers of the latter. Also, we did not have information on when the initial symptoms of AAV appeared and there might have been a lag in time between initial symptoms until the actual diagnosis was settled, which could have confounded data regarding seasonal variation. The sensitivity of the diagnostic codes was not assessed; hence, although the incidence of AAV found in our study was comparable to previous findings, we may not have identified all cases in the study period. Nonetheless, as we were able to prove that >93% of the study population identified in the Northern Region of Denmark at least had one ANCA test taken, and that >90% of our primary study population (identified by ICD10 codes) was confirmed by including a new and validated set of inclusion criteria, with subsequent reproducibility of key findings, we strongly believe the AAV cohort applied in the original manuscript is reliable and suitable for the study purpose. While hypertension and diabetes did not add significantly to the statistical model, there was clearly an increase in both attributes throughout the three time periods. This may pertain to altered coding habits during the years; however, specifically it is related to altered treatment and diagnostic guidelines that were implemented during the same period [51, 52], as we included treatment with two or more antihypertensive medications and oral antidiabetic as proxies for these diseases, respectively. Also, accuracy of hospital codes denoting ESRD may have improved over time; such variation was, however, not evident in our results, with risk of ESRD decreasing over time.

CONCLUSIONS

There was an increase in incidence of AAV between 2000 and 2015 in Denmark. This may have been influenced by greater use of ANCA testing, which is reflected in patients being diagnosed at an earlier stage, more elderly patients being diagnosed and patients having better outcomes in terms of survival and ESRD.

SUPPLEMENTARY DATA

Supplementary data are available at ndt online.

FUNDING

K.E.N.K. was funded privately by Helen and Ejnar Bjørnows Foundation and Knud Højgaards Foundation.

AUTHORS’ CONTRIBUTIONS

K.E.N.-K., W.S., M.E., N.C., D.J. and M.M. were responsible for concept and design. K.E.N.-K., W.S., M.E., N.C., D.J., M.M., H.D., J.W.G., E.K., P.I. and C.T.-P. participated in analysis and interpretation of data. K.E.N.-K. drafted the initial manuscript and W.S., M.E., N.C., D.J., M.M., H.D., J.W.G., E.K., P.I. and C.T.-P. were responsible for the revision process. All authors provided intellectual content of critical importance to the work described and all authors finally approved the version to be published.

CONFLICT OF INTEREST STATEMENT

D.J. has received research grants from Chemocentryx, GSK, Roche/Genentech and Sanofi-Genzyme. He has received consultancy fees from Astra-Zeneca, Boehringer-Ingelheim, Chemocentryx, Chugai, GSK, Infla-RX, Insmed and Takeda.

REFERENCES