RATIONALE
Although osteoarthritis (OA) is highly prevalent, there are no strong evidences for disease-modifying interventions. However, there are many symptom relieving treatments available and arthroplasty is a very useful treatment for end-stage disease.
In addition, OA is a heterogeneous disease and not all patients progress clinically or radiologically.
In order to target treatments to patients with OA, or recruit patients into clinical trials, it is important to know their risk of progression and the phenotypic features driving the disease. To facilitate this, we need to identify novel biomarkers and clinical predictors of disease progression and validate them in large existing cohorts.
OVERARCHING AIMS
This project will develop a unique European collaboration with access to the largest collection of cohorts (Work Stream 1) and healthcare databases (Work Stream 2) worldwide and all of the expertise to accurately phenotype patients with OA in terms of their risk of disease progression, their phenotypic features, and the potential to produce a prediction model for individual patients. This will allow targeting of treatments and optimal randomization into clinical trials. It will also provide the perfect platform for the testing and validation of novel and existent biomarkers.
SPECIFIC AIMS (WS2)
1.To develop and validate algorithms to accurately predict risk of primary arthroplasty in newly diagnosed cases of hip and knee OA, using routinely collected data and harmonised across different European healthcare databases
2.To explore the potential for improved accuracy of risk prediction afforded by (a) updating individual risk prediction with post-diagnosis events and trajectories, (b) incorporating additional, novel prognostic indicators identified from Workstream 1
3.To explore the effectiveness of pharmacological interventions on the need for lower limb (hip or knee) arthroplasty.
METHODS (WS2)
-Data sources: The available eligible population from 5 participating databases from Denmark, Sweden, the UK, Catalonia/Spain, and the Netherlands (see co-app contributions in section 1 and appendix V in the full study protocol) will be used (estimated combined total population aged ?40 years of >10 million subjects).
-Study population: patients aged 40 or older with a GP (CPRD, SIDIAP, Netherlands) or hospital (Denmark, Sweden) diagnosis of hip or knee osteoarthritis will be identified as eligible participants.
-Study outcome: total knee/hip arthroplasty will be the primary study outcome, as a surrogate for joint failure. Time from GP/hospital diagnosis of knee/hip osteoarthritis to surgery will be modelled.
-Statistics:
Objective 1. Derivation and validation of models to predict future likelihood of primary arthroplasty: The primary outcome will be primary total hip or knee arthroplasty identified from electronic healthcare databases and linked registries. Cases of incident hip and knee OA will be identified after 1 January 2006. Separate models will be produced for hip OA and for knee OA. Candidate prognostic indicators will be identified by systematic literature search and expert consensus. Models will be fitted and internally validated using bootstrapping in CPRD, which is the largest, most comprehensive database, followed by external validation and recalibration in the other European databases.
Objective 2. Improving the predictive performance of models: We will explore two novel avenues for improving the accuracy of prediction of the above models: (a) updating the model with post-diagnosis events and trajectories, e.g. change in BMI, escalation of prescribed analgesia. (b) Evaluating the potential contribution of additional marginally informative markers of osteoarthritis progression identified in Workstream 1 using direct empirical evaluation and exploring the potential to use simulation.
Objective 3. Comparative effectiveness research: exploring interventions that modify the likelihood of arthroplasty. The models detailed above, provide a potential common platform for pharmaco-epidemiological studies, specifically comparative effectiveness research, investigating the effects (beneficial or detrimental) of prescription drugs available in all the participating countries on likelihood of arthroplasty. In this study we will use propensity score approaches to minimize confounding by indication. Additionally, we will use an incident user design to estimate the effect of prescribed drugs (eg. bisphosphonates, strontium) on the risk of arthroplasty.
-Sample Size Considerations: Assuming a conservative risk of arthroplasty of 10% at 5 years, even the smaller databases (pop = 0.6m) will have sufficient power to enable the fitting of >100 predictors in the predictive models.