American College of Rheumatology/European League Against Rheumatism (ACR/EULAR) Diagnostic and Classification Criteria for Primary Systemic Vasculitis

The systemic vasculitides are a group of uncommon but important diseases whose prognosis has improved dramatically with the use of immunosuppressive therapy. However, long-term morbidity from recurrent disease flares, low-grade grumbling disease and/or accumulating damage from previous disease activity or drug therapy now characterise the long-term outlook for patients with vasculitis. There remains major controversy, and incompatibility between the ANCA-associated vasculitides: granulomatosis with polyangiitis (Wegener's), microscopic polyangiitis, and Churg Strauss Syndrome, as well as polyarteritis nodosa in the current classification criteria and disease definitions. Importantly, there are no diagnostic criteria for any of the primary systemic vasculitides. We propose to improve existing classification criteria for the primary systemic vasculitides. As a starting point will include the following diseases: granulomatosis with polyangiitis (Wegener's) (GPA), microscopic polyangiitis (MPA), Churg Strauss syndrome (CSS), polyarteritis nodosa (PAN), giant cell arteritis (GCA) and Takayasu arteritis (TAK). We propose to develop and validate classification and diagnostic criteria for primary systemic vasculitis using the guidelines suggested by the Classification and Response Criteria Subcommittee of the American College of Rheumatology Committee on Quality Measures. For all patients, a detailed medical history, physical examination, laboratory tests (including ANCA), radiology (including angiography), biopsy results, treatment, Birmingham Vasculitis Activity Score (BVAS)version 3, Vasculitis Damage Index (VDI), will be collected. The exact list of items to be recorded will be determined by the expert panel at the start of the study. Classification criteria We will study a minimum of 100 patients (new and existing patients) prospectively within each currently defined disease category (GPA, CSS, MPA, PAN, GCA, TAK) for the development of the classification criteria. We anticipate the need to recruit 130 patients to account for misdiagnosis and dropout to achieve the target of 100 with the confirmed reference diagnosis. This will include patients that have vasculitis which are assumed to be related to ANCA but do not fulfil the current definitions of any of the diseases, and patients with large vessel vasculitis which do not fulfil current definition for GCA or TAK. Therefore new categories of disease may be created as part of this process and some of the current disease categories may be changed to include or exclude certain patients. The other diseases will be the controls. The same minimum number of patients will be used to validate the criteria. The 1st 100 patients with a formal reference diagnosis that are recruited for each disease will be used for development of the classification criteria; the next 100 consecutive patients recruited with a confirmed reference diagnosis for each disease will be used to validate the criteria. Again we anticipate the need to recruit 130 patients to account for misdiagnosis and dropout to achieve the 100 target. The majority of cases included will be the same as that used for the development of the diagnostic criteria. In the absence of an established gold standard, we propose to develop a reference standard. Clinical vignettes using clustering of clinical features and investigations will be constructed from actual cases by the steering group. An expert panel will then be asked to classify each vignette. Hypothetical changes will then be made to components of each clinical vignette and the expert panel will be asked to re classify the case. This process will be repeated multiple times in an attempt to determine what key clinical feature influence the expert panel to change the diagnosis. Using this data driven process, a construct of important clinical features for each disease will be determined by the expert panel. Using this new construct, patients will be classified by the expert panel. This will form the reference standard against which the new criteria will be tested. Diagnostic Criteria We propose to develop and validate diagnostic criteria for primary systemic vasculitis. Based on current disease categories we will include GPA, MPA, CSS, PAN, GCA and TAK (but this may change depending on whether new categories are created or existing categories merged as part of the classification criteria component). For the development of diagnostic criteria, we will study a minimum of 100 patients (will require approx 130 patients to allow for dropout and misdiagnosis) for each disease category. Assuming 6 disease categories, the majority of these 780 patients will have already been identified from the classification criteria component of the study and will be re used for the development and validation of diagnostic criteria. However, for the diagnostic criteria to be clinically relevant we will only include patients that are seen at the time of 1st presentation, therefore not all the 780 patients recruited for the classification criteria section of the study will be suitable, and we will need to recruit additional new patients for each of the types of vasculitis being studied. We will use a minimum of 400 context specific controls (patients that don't have vasculitis) for AAV and PAN that will cover the spectrum of different disease presentations and severity. In addition, we will recruit a minimum of 100 context specific controls for GCA and a similar number for TAK. Different control populations are needed for AAV, GCA and TAK as they have significantly different clinical presentations. In a similar manner to cases, we will recruit 30% more patients than the minimum required to account for misdiagnosis and drop out. The same minimum number of cases and controls will be needed to validate the criteria. The first half of the patients recruited would be used to develop the criteria, and the 2nd half to validate the criteria. We will allow inclusion of patients from previously studied prospective cohorts that meet all the appropriate inclusion / exclusion criteria and have had all the appropriate clinical information and mandatory investigation (to be defined later) recorded at time of their first presentation. This is to facilitate the recruitment of sufficient patients with PAN, CSS and TAK which are rare conditions. Statistical analysis We will follow the ACR recommended statistical methods for creating the classification criteria. Patients will have been classified into the different types of vasculitis according to the proposed EULAR/ACR schema by the expert panel or as a vasculitis mimic. The outcomes of interest are binary variables indicating whether or not a patient has been classified as having a particular type of vasculitis, such as GPA, MPA, etc. For each outcome, multivariable logistic regression modeling will be used to identify predictors of outcome based on the list of potential predictor variables described earlier. We will also explore the use of Classification And Regression Tree (CART) analysis. This is a tree-building technique ideally suited to the generation of clinical decision rules. Unlike conventional regression methods, patients are partitioned ("split") into different groups based on an exhaustive search of all possible predictor variables. The advantage of CART analysis over conventional methods is that it is non-parametric, so no assumptions are made about the underlying distribution of predictor variables. CART can handle many hundreds of possible predictor variables and can uncover complex interactions between predictors which may be difficult or impossible to uncover using traditional multivariate techniques that can suffer from model over fitting. In addition, clinicians generally do not think in terms of probability but rather in terms of categories, such as low versus high risk. Clinical decision rules generated using CART analysis are more likely to make clinical sense, and hence more likely to be followed in clinical practice. Once the best items are identified, the expert panel will decide on the best short list of items to be included in each criteria and also choose the most appropriate decision tree. This will provide the best content validity. The statistical methods to be used for diagnostic criteria will be very similar to that used for the classification criteria. The binary outcome for analysis is whether the person is a case or control (without vasculitis). We repeat the analyses for each of each type of vasculitis e.g. WG versus controls, then CSS versus controls etc