Leveraging a national patient registry to explore relapse in ANCA associated vasculitis
Citation:
Scott, Jennifer Orna, Leveraging a national patient registry to explore relapse in ANCA associated vasculitis, Trinity College Dublin, School of Medicine, Clinical Medicine, 2023Abstract:
BACKGROUND
ANCA associated vasculitis (AAV) is an archetypal rare autoimmune disease, resulting in immune-mediated organ damage. It has a relapsing-remitting course, resulting in cumulative morbidity and mortality. Toxic immunosuppression used to induce and maintain remission is a double-edged sword; >80% experience adverse events. A key unmet need is personalisation of this treatment. Identifying and predicting relapse are key enablers of this. Relapse is defined by the Birmingham Vasculitis Activity Score (BVAS v3) >0, particularly in the clinical trial setting. However, this metric may be missing or incorrectly scored in real world registry data, resulting in incomplete or inaccurate ascertainment of this key outcome. Furthermore, the precise triggers of relapse are unknown, limiting prediction. Evidence suggests a complex interplay of polygenic genetic susceptibility, epigenetic influences, and environmental triggers ? the latter receiving the least attention to date. High quality registry data is a prerequisite for meaningful research in this area.
AIMS
We sought to:
i. Build a seamless, transparent data pre-processing pipeline to enable real-world research into relapse in AAV, with optimisation of data quality (DQ) in the Rare Kidney Disease (RKD) registry a key focus.
ii. Create a computerised algorithm to automate the identification of relapse in AAV, in real-world registry-based trials.
iii. Explore the long-term outcomes of Irish patients with AAV recruited to the RKD registry, and investigate potential prognostic factors, to enable comparisons with international benchmarks.
iv. Conduct a systematic review of the role of environmental influences in AAV.
v. Evaluate the association of ultraviolet B (UVB) radiation with relapse in AAV.
METHODS
The RKD Registry is an Irish national longitudinal, multi-centre, cohort study, founded in 2012, which includes 663 patients with definite AAV.
i. Registry design was improved in 6 key DQ domains: uniqueness, consistency, completeness, accuracy, stability, and timeliness. Standard operating procedures were implemented to streamline data processes. Annotated, version-controlled R code was developed to automate a reproducible flow of data from raw- to usable-formats. The benefits and challenges of integrating a smartphone application for capture of real-time patient telemetry were also explored.
ii. Patient encounters (N=3387) were summarised using six objective variables: diagnostic histopathology, change in ANCA level, suggestive blood/urine tests, suggestive imaging, immunosuppressive (IS) status at the time of the encounter and the change of this IS in response (`IS response?). The probability of relapse was independently adjudicated for each encounter, by an expert committee (ground truth). A computable phenotype definition (CPD), with an embedded multi-level model (MLM), was developed and internally validated, to assign an individualised probability of relapse for each encounter. A Shiny R web application was developed to implement the final algorithm.
iii. Survival analysis and cox proportional hazards models were used to explore the outcomes of 397 prospectively recruited patients, including mortality, end-stage kidney disease (ESKD) and relapse.
iv. A systematic scoping review was performed to describe what is currently known about the effect of the environment on AAV disease activity.
v. An n-of-1 study design was employed to examine the effect of UVB-derived exposure variables on relapse risk, using MLMs. R studio was used for all analyses.
RESULTS
i. A robust data ecosystem was developed, with tangible improvement in DQ measured over time, to support meaningful registry-based research, described in this thesis and beyond.
ii. The CPD achieved an F1-Score (an harmonic mean of precision and recall, equivalent to accuracy in an imbalanced dataset) of 0.85, with sensitivity and specificity of 0.91 and 0.95, respectively (for the complete model). Alternative models, with corresponding performance metrics, were derived for instances where 1 or more variables were missing. Where `IS response? was missing, `suggestive bloods/urine? with at least either `ANCA level? or `suggestive imaging? was required to achieve an accuracy as good as the gold standard BVAS >0 relapse definition.
iii. Cumulative five-year patient, renal and relapse-free survival were 77%, 79% and 64%, respectively, in line with international norms. Baseline renal dysfunction and the burden of adverse events were independent predictors of death overall. Baseline severity of renal insufficiency, urine soluble CD163 (usCD163) and ?sclerotic? Berden histological class were key determinants of ESKD risk. ANCA serotype, cardiovascular involvement and baseline renal function were independent risk factors for relapse.
iv. Average vitamin D-UVB (UVB at wavelengths specific for vitamin D synthesis) was negatively correlated (0.82, 95% CI 0.70-0.99) with relapse risk, with a stronger effect when restricting to winter measurements (0.71, 95% CI 0.57-0.89).
CONCLUSIONS
i. Standardised design, systems and procedures are paramount to optimal data management in patient registries, which is a key component of rare disease research.
ii. In settings where accurate BVAS may not be available, our CPD accurately quantifies the individualised probability of AAV relapse using objective, readily accessible registry data.
iii. Long-term outcomes of Irish patients with AAV are comparable to other reported international series. Our results emphasise the need for personalisation of immunosuppression, to limit treatment toxicity, particularly in those with advanced age and renal insufficiency.
iv. Prolonged low winter ambient UVB is associated with an increased risk of relapse in AAV, across all phenotypes. Given a person?s vitamin D status is largely dependent on their UVB exposure, the next step will be to investigate if avoidance of vitamin D deficiency can reduce relapse propensity in AAV.
v. The work described in this thesis formed the nidus of a large EU grant, to create a multi-modal prediction model, to stratify relapse risk in AAV, to enable personalised prescribing.
Sponsor
Grant Number
The Irish Clinical Academic Training (ICAT) Programme, supported by the Wellcome Trust and the Health Research Board (Grant Number 203930/B/16/Z), The Health Service Executive, National Doctors Training and Planning and the Health and Social Care Research and Development Division, Northern Ireland.
203930/B/16/Z
Description:
APPROVED
Author: Scott, Jennifer Orna
Advisor:
Little, MarkWhite, Arthur
Ng, James
Walsh, Cathal
Kelleher, John
Aslett, Louis
Publisher:
Trinity College Dublin. School of Medicine. Discipline of Clinical MedicineType of material:
ThesisCollections
Availability:
Full text availableMetadata
Show full item recordLicences: