To determine the association between characteristics at diagnosis and the time to first relapse in a large cohort of patients with antineutrophil cytoplasmic antibody–associated vasculitis (AAV).
To determine the association between characteristics at diagnosis and the time to first relapse in a large cohort of patients with antineutrophil cytoplasmic antibody–associated vasculitis (AAV).
We studied long-term followup data from 4 clinical trials that included newly diagnosed patients with a broad spectrum of AAV severity and manifestations. Patient and disease characteristics at baseline were used in competing risk regression models with relapse as the event of interest and death as the competing event.
We assessed 535 patients with 1,804 patient-years at risk of relapse. At diagnosis, the median age was 60.7 years (interquartile range [IQR] 48.8–69.1 years), 284 patients (53%) had granulomatosis with polyangiitis (Wegener's), and the median creatinine level was 203 μmoles/liter (IQR 97–498). A total of 201 patients (38%) experienced a relapse and 133 patients (25%) died, 96 of whom had not had prior relapse. Anti–proteinase 3 antibodies (subhazard ratio [sHR] 1.62 [95% confidence interval 1.39–1.89]) and cardiovascular involvement (sHR 1.59 [95% confidence interval 1.07–2.37]) were independently associated with a higher risk of relapse. Compared with patients with a creatinine level ≤100 μmoles/liter, patients with higher creatinine levels had a lower risk of relapse (sHR 0.81 [95% confidence interval 0.77–0.85] for a creatinine level of 101–200 μmoles/liter; sHR 0.39 [95% confidence interval 0.22–0.69] for a creatinine level >200 μmoles/liter).
Relapse of disease is common for patients with AAV. A creatinine level >200 μmoles/liter at the time of diagnosis is strongly associated with a reduced risk of relapse and may help guide monitoring and treatment of patients with AAV.
Granulomatosis with polyangiitis (Wegener's) (GPA), microscopic polyangiitis (MPA), and renal limited vasculitis are among the most common small-vessel vasculitides in adults. They are characterized by the presence of antineutrophil cytoplasmic antibodies (ANCAs) and are frequently grouped together as ANCA-associated vasculitis (AAV) (1, 2). Early in the 20th century, AAV was fatal (3, 4). The introduction of immunosuppression with alkylating agents and glucocorticoids, widespread availability of ANCA testing, and renal replacement therapy have dramatically reduced mortality rates. AAV is now a chronic disease characterized by periods of remission and relapse.
The prevention of relapses is important given their effects on the patient's quality of life and the potential to lead to disability or death (5, 6). However, treatment to prevent relapses carries a burden of cumulative drug toxicity that may also contribute to reduced quality of life, disability, and potentially early mortality. Developing monitoring and treatment strategies to prevent relapses requires an understanding of which patients are at high risk. Previous studies on relapses have had limited success in identifying patients at risk. Furthermore, most studies in AAV tend to be single-center cohorts or from limited geographic areas and are often retrospective in design (5–7). In particular, due to the relative rarity of the disease and its protean manifestations, few studies assess both the full spectrum of manifestations and severity of AAV in appreciable numbers. It therefore remains uncertain whether patients who differ with respect to disease severity and/or particular manifestations of AAV have different risk profiles.
Given the current limitations of identified risk factors for relapse in AAV, we assessed the independent association between patient and disease characteristics at the time of diagnosis and the time to first relapse. We assessed these associations using data from 4 clinical trials conducted across Europe with extended followup periods.
The European Vasculitis Study Group conducted 4 trials that enrolled patients from 70 hospitals in 15 countries between 1995 and 2002 (8–11). The trials were conducted according to the 1964 Declaration of Helsinki and subsequent amendments. All patients were newly diagnosed as having AAV (GPA, MPA, or renal limited vasculitis) using definitions of AAV adapted from the Chapel Hill Consensus Conference definitions (1). One trial enrolled patients with early systemic AAV, 2 trials enrolled patients with generalized AAV, and 1 trial enrolled patients with severe AAV. The individual trial eligibility criteria are summarized in Table 1. The NORAM (efficacy of methotrexate versus cyclophosphamide in the treatment of early non-renal Wegener's granulomatosis) (9), CYCAZAREM (randomized trial of cyclophosphamide versus azathioprine during remission in ANCA-positive systemic vasculitis) (8), and CYCLOPS (randomized trial of daily oral versus pulse cyclophosphamide as therapy for ANCA-associated systemic vasculitis) (10) trials compared low toxicity regimens (methotrexate, azathioprine, and intravenous cyclophosphamide) with standard therapy (oral cyclophosphamide), while the MEPEX (randomized trial of adjunctive therapy for severe glomerulonephritis in ANCA-associated systemic vasculitis: plasma exchange versus intravenous methylprednisolone) trial compared 2 adjunctive treatments (intravenous glucocorticoids and plasma exchange) (11).
|Trial (ref.)||Included disease stage||Creatinine, μmoles/liter||Induction treatment||Maintenance treatment|
|NORAM (9)||Early systemic||<150||MTX vs. oral CYC||MTX vs. oral CYC|
|CYCAZAREM (8)||Generalized||<500||Oral CYC||Oral CYC vs. AZA|
|CYCLOPS (10)||Generalized||<500||IV CYC vs. oral CYC||AZA|
|MEPEX (11)||Severe||≥500||PE + oral CYC vs. IVMP + oral CYC||AZA|
Patients were assessed at baseline for manifestations of AAV in each organ system using the Birmingham Vasculitis Activity Score (BVAS), an instrument with 9 domains (general, cutaneous, mucous membranes/eyes, ears/nose/throat, chest, cardiovascular, abdominal, renal, and nervous system) (1). The BVAS produces a summary score, and each domain was also classified as actively involved or not on the basis of ≥1 item or no items present at baseline. Patients were followed up within the NORAM, CYCAZAREM, and CYCLOPS trials for 18 months and within the MEPEX trial for 12 months. Patients were then followed up under routine clinical care in an extended followup study with summary data on relapses and vital status recorded between 2005 and 2008 corresponding to followup between 36 and 132 months of followup for survivors.
Relapses were defined as new or worsened manifestations of AAV that were also believed to require a change in therapy. Death was defined as death due to any cause. Patient data were censored at the time of relapse, death, or last followup visit, whichever occurred first.
Summary data are presented as the median and interquartile range (IQR) for continuous variables and as proportions for categorical data. Baseline characteristics were compared between trials using the Mann-Whitney U test for continuous variables and Fisher's exact test for dichotomous data. Time to first relapse is displayed graphically either as a 1 − Kaplan-Meier estimate (censored at the time of death) or as cumulative incidence curves (death as a competing event).
Models used to identify independent risk factors were created by selecting baseline variables that represent patient characteristics and disease characteristics. Patient characteristics included age and sex. Disease characteristics included those that corresponded to disease severity (serum creatinine level, summary BVAS, and C-reactive protein [CRP] level) and those that corresponded to clinical phenotype (diagnostic subtype [GPA compared to non-GPA], anti–proteinase 3 [anti-PR3] antibodies or PR3 cytoplasmic ANCA immunofluorescence, and individual organ system involvement). Creatinine level was categorized as ≤100 μmoles/liter, 101–200 μmoles/liter, and >200 μmoles/liter based on initial modeling experience demonstrating a nonlinear relationship with hazard of relapse.
Due to the correlation between organ system involvement, diagnostic subtype (GPA or MPA), and summary BVAS (i.e., both diagnostic subtype and BVAS are determined by type and number of organs involved at baseline), models using organ system involvement (model 1) were fit separately from models using diagnostic subtype and BVAS (model 2). Renal system involvement was excluded as a baseline variable in model 1 due to its dependence on the serum creatinine level.
We constructed competing risk models in which death was treated as a competing event because variables that represent disease severity that are associated with mortality may also be associated with relapse (13). Patients with risk factors that increase mortality may therefore be censored early due to death in traditional time-to-event analysis, and the association of those risk factors with relapse may be obscured. Competing risk models jointly estimate the associations of each factor with the competing risk (death) and the event of interest (relapse). The proportional hazards assumption was assessed by visually inspecting log–log plots and by assessing each variable as a time-varying covariate. The competing risk models provide subhazard ratios (sHR) as a measure of association between each characteristic and relapse (i.e., the HR for relapse after accounting for the joint probability of death). The nonindependence of patients enrolled in the same trial (i.e., associations between characteristics and outcomes for patients within the same trial are correlated) was accounted for by using robust variance estimates based on clustering by trial. This procedure results in more valid variance estimates for associations without altering the point estimates. All candidate predictor variables were fit and retained in all models (forced entry), and no interaction terms were included.
We used multiple imputation with chained equations to account for missing covariate data with 10 imputation data sets (14, 15). The following sensitivity analyses were conducted: use of complete cases with no imputation methods, and analysis using standard Cox regression rather than competing risk regression. Sensitivity analyses of individual trials and analyses excluding the NORAM and MEPEX trial data were also conducted to ensure that inclusion of trials that enrolled patients with the extremes of disease severity was not primarily responsible for the observed associations.
Associations are presented as point estimates with 95% confidence intervals (95% CIs). P values less than 0.05 were considered significant with no adjustment made for multiple comparisons. All analyses were performed using Stata MP software, version 11.
A total of 535 patients were enrolled in the 4 trials. Seventeen percent of patients had at least 1 baseline variable missing, usually the CRP level. At trial entry, the median age was 60.7 years (IQR 48.8–69.1 years), and the median creatinine level was 203 μmoles/liter (IQR 97–498) (Table 2). Fifty-three percent of patients had GPA, and 56% were anti-PR3 positive. Trials differed significantly in terms of most variables as expected by their entrance criteria, which stratified patients to trials on the basis of disease severity.
|All trials (n = 535)||NORAM (n = 95)||CYCAZAREM (n = 155)||CYCLOPS (n = 148)||MEPEX (n = 137)||P|
|Age, median (IQR) years||60.7 (48.8–69.1)||53.3 (39.7–61.2)||57.6 (47.3–67.6)||60.7 (48.5–68.5)||66.5 (59.6–71)||<0.001|
|Creatinine, median (IQR) μmoles/liter||203 (97–498)||82 (71–92)||172 (97–275)||182 (123–302)||738 (601–918)||<0.001|
|CRP, median (IQR) mg/liter||54 (18–118)||46 (20–125)||47 (13–111)||35 (13–101)||79 (35–158)||<0.001|
|BVAS, median (IQR)||18 (12–23)||13 (9–20)||17 (10–24)||20 (15–24)||17 (15–24)||<0.001|
|Organ system involvement|
Patients accumulated 1,804 patient-years of time at risk (i.e., from time of entry to time of first relapse, death, or end of followup), during which time 201 patients (38%) had at least 1 relapse and 133 patients (25%) died, 96 of whom had not had prior relapse. The time to first relapse is demonstrated in Figure 1, both as a cumulative incidence curve in which death is a competing event and as a traditional 1 − Kaplan-Meier estimator curve censoring for death. The 2 different methods result in an estimated 9% lower incidence of relapse at 5 years using competing risk analysis.
Results from the competing risk regression are summarized in Table 3. There was no evidence of violation of the proportional hazards assumption in either model. In model 1, anti-PR3 positivity (sHR 1.62 [95% CI 1.39–1.89], P < 0.001) and cardiovascular involvement (sHR 1.59 [95% CI 1.07–2.37], P = 0.02) were independently associated with time to first relapse. Higher serum creatinine level was independently associated with a decreased risk of relapse. Compared to patients with a creatinine level ≤100 μmoles/liter, patients with a creatinine level of 101–200 μmoles/liter had an sHR of 0.81 (95% CI 0.77–0.85) (P < 0.001), and patients with a creatinine level >200 μmoles/liter had an sHR of 0.39 (95% CI 0.22–0.69) (P = 0.001) (Table 3 and Figure 2). Model 2, which included BVAS and diagnostic subtype but not individual organ involvement terms, demonstrated associations between relapse and anti-PR3 positivity and serum creatinine level that were similar to those in model 1. Diagnosis of GPA (sHR 1.44 [95% CI 1.06–1.94], P = 0.02) was also associated with a higher risk of relapse. In neither model 1 nor model 2 were general markers of baseline disease severity (CRP level, BVAS) independently associated with time to first relapse.
|Model 1||Model 2|
|sHR (95% CI)||P||sHR (95% CI)||P|
|Age, per year||1.00 (0.99–1.01)||0.72||1.00 (0.99–1.01)||0.97|
|Female||0.85 (0.65–1.11)||0.24||0.88 (0.70–1.10)||0.25|
|Anti-PR3 positive||1.62 (1.39–1.89)||<0.001||1.57 (1.12–2.19)||0.008|
|101–200 μmoles/liter||0.81 (0.77–0.85)||<0.001||0.77 (0.68–0.87)||<0.001|
|>200 μmoles/liter||0.39 (0.22–0.69)||0.001||0.39 (0.24–0.64)||<0.001|
|Second quartile||1.07 (0.69–1.66)||0.77||1.02 (0.65–1.6)||0.93|
|Third quartile||0.87 (0.67–1.14)||0.32||0.86 (0.56–1.3)||0.46|
|Fourth quartile||1.01 (0.65–1.57)||0.97||0.97 (0.53–1.78)||0.91|
|Second quartile||NA||NA||0.89 (0.57–1.38)||0.60|
|Third quartile||NA||NA||0.89 (0.66–1.20)||0.44|
|Fourth quartile||NA||NA||1.23 (0.8–1.9)||0.35|
|Organ system involvement|
|Mucous membranes/eyes||1.39 (0.95–2.05)||0.09||NA||NA|
Sensitivity analyses using only complete cases (i.e., excluding all cases with at least 1 missing variable) resulted in no significant changes in the results. Standard Cox regression models also demonstrated results that were qualitatively similar to those in the competing risk model, although age was associated with a small increase in risk of relapse, while a creatinine level >200 μmoles/liter was associated with less of a protective effect (further information is available at www.vasculitis.org/default.aspx?code=04). Effect estimates, including those for serum creatinine level, were consistent between individual trials and when the MEPEX and NORAM trial data were excluded from pooled analyses.
Classifying patients by the number of risk factors present at diagnosis demonstrated that only 95 of the 383 patients (25%) with no risk factors (i.e., no cardiovascular involvement, not anti-PR3 positive, and a serum creatinine level >100 μmoles/liter) had a relapse at 5 years. In contrast, 10 of 37 patients (27%) with 1 risk factor and 74 of 125 patients (59%) with ≥2 risk factors had a relapse (Figure 3).
We studied the relationship between factors at the time of diagnosis of AAV and the outcome of relapse in an international cohort of 535 patients spanning the full spectrum of manifestations and severity of disease. Treatment with conventional immunosuppression and glucocorticoids results in an overall risk of relapse of 38% at 5 years. We found that anti-PR3 antibody positivity and cardiovascular involvement were independent predictors of relapse. Further, we found that patients with higher serum creatinine levels were at a substantially lower risk of relapse.
Our findings are consistent with those of previous studies of relapse in AAV (7, 16–19). Previous studies have identified anti-PR3 positivity (7, 19) and GPA (16, 20) as predictors of relapse. Further, the magnitudes of these associations between ANCAs and relapse are similar between our study (HRs of ∼1.5 for each risk factor in our study) and those of others (HRs of ∼1.7) (7, 19). However, unlike other studies, we found that cardiovascular involvement was associated with a higher risk of relapse, and poor renal function was associated with a lower risk of relapse. Previous studies may have missed this association because of any one of several reasons. For example, other studies may have had limited power (i.e., few patients with cardiovascular manifestations were included) or cardiovascular involvement was not assessed. Alternatively, perhaps our finding is due to random error. Further, cardiovascular manifestations are difficult to diagnose and to attribute correctly to AAV, which creates a large potential for misclassification error. Most of these issues would reduce the probability of detecting an association between cardiovascular manifestations and risk of relapse. Although cardiovascular manifestations occurred in only 5.7% of our cohort and are therefore relevant to only a small number of patients, they can be life-threatening; given our findings that they may be associated with relapse, careful consideration of their presence at diagnosis may be of considerable clinical importance.
We also documented a strong association between worse renal function and lower risk of relapse. A previous small study also found that reduced renal function was independently associated with a lower risk of relapse, while several others have either found no association or have not reported an association between renal function and relapse (7, 21–23). Other studies demonstrated a reduced relapse rate only for patients with end-stage renal disease. Differences between our results and those of others may be due to the broad spectrum of disease used in our study, which allowed us to discriminate between patients with normal renal function and those with only moderate levels of impairment. Also, the relatively large number of patients and events in our study allowed us greater power to detect associations and avoid model overfitting. Finally, the use of a competing risk framework may have allowed us to detect this association free of the confounding effects of renal function on mortality.
One may assume that aggressive AAV, such as that accompanied by severe renal dysfunction, is likely to relapse. The paradoxical finding that more severe presentations may be associated with a lower risk of relapse may be due to the state of immunosuppression caused by renal dysfunction (24–27). Other severe manifestations, such as cardiovascular manifestations, are associated with an increased risk of relapse. Thus, relapse risk is more likely related to which organs are affected than to the overall severity or aggressiveness of the vasculitis. This finding may have direct clinical importance. Patients with severe renal dysfunction are at the highest risk of death, and many of these deaths are due to infection and adverse sequelae from treatment (28). Given that these patients are at a lower risk of relapse and at high risk of death from treatment-related adverse events (11), they may benefit from relatively less intense immunosuppression after an initial response to therapy.
Our study has several notable strengths. It includes the largest number of patients in any study of relapses in AAV that we are aware of. Our patient population also spans many European centers, and our study considers the full range of clinical manifestations and severity of disease at diagnosis and considers all patients from the time of diagnosis forward. Finally, we made use of advanced statistical techniques such as competing risk analysis and multiple imputation. The competing risk framework is sensible in this disease and has added new insights into the assessment of baseline risk factors. The use of multiple imputation allows the use of all the data, which may improve the precision of estimates and reduce bias created by missing data (29).
Our results must also be viewed in terms of their limitations. Although our study is the largest of its kind, it is still relatively small compared with risk factor analyses in other, more common chronic diseases (e.g., cardiovascular disease). Despite this, we have many relapse events which allow us to fit stable regression models (30). We are therefore more at risk of missing moderate-to-small associations than we are of finding spurious associations. Second, our study is also not a true inception cohort because the patient population is drawn from clinical trials. However, the selected trials purposefully represent the full spectrum of AAV, and centers typically enrolled a high percentage of all available patients with AAV in the trial. Although the absolute relapse rate may not be generalizable, the associations are likely generalizable to the wider population of patients with AAV because the biology of these relationships is unlikely to differ between patients in an inception cohort and patients enrolled in clinical trials.
Third, the use of different clinical trials also introduces heterogeneous inclusion criteria and treatments. We have ensured that effect estimates are consistent across trials, adjusted for differing treatments, and used statistical methods to take into account the clustered nature of the data within trials. Further, the trials were designed to enroll patients from the entire population of patients with AAV, and thus their grouping together much more closely approximates the population than do any trials individually.
Finally, the assessment of risk of relapse in clinical practice is a longitudinal process. Our study only makes use of clinical information available at baseline. Further studies are required to understand the best integration of longitudinal data and the potential added benefits of novel biomarkers into risk assessment.
In summary, we found that patients who are anti-PR3 positive, have cardiovascular involvement, and do not have renal impairment at the time of diagnosis are at the highest risk of relapse. Identifying patients at high risk of relapse may aid monitoring strategies and weighing the relative benefits of different treatment strategies. Further research assessing changes in baseline parameters may improve our ability to predict which patients will have relapse of their disease.
All authors were involved in drafting the article or revising it critically for important intellectual content, and all authors approved the final version to be published. Dr. Walsh had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.
Study conception and design. Walsh, Jayne.
Acquisition of data. Flossmann, Berden, Westman, Höglund, Stegeman, Jayne.
Analysis and interpretation of data. Walsh, Jayne.