Search

The goal of WP3 “Personalized Medicine” is to establish a standardized process to enable personalized decision-making that can be utilized for multiple outcomes of interest and can be applied to observational healthcare data from any patient subpopulation. In the first report (D3.2) on the implementation of the analytical pipeline for personalized medicine, we introduced the analytical pipelines for Patient-Level Prediction and Population-Level Effect Estimation. Furthermore, we discussed our initial work for the development of a pipeline for Risk Stratified Effect Estimation to assess heterogeneity of treatment effect. In the second report (D3.4) we provided an update on the work done in the second year, including heterogeneity of treatment effect (HTE), and disease trajectory analyses. That deliverable included an overview of use cases in which the analytical pipelines have been applied and described the advances made in methodological research, the start of a natural language processing pipeline, and work done to develop a pipeline for disease trajectories. In the third report (D3.6) we provided further updates to the methodological research and the introduction of new tools including exhaustive search and frequent pattern mining, the use of unstructured data in predictive analytics as well s publishing a systematic review on clinical prediction modelling. In this final report (D3.8) we introduce new additional analytical tools and provide a further update on the methodological research and clinical applications. This work falls under Task 3.2. “Development of an integrated patient-level prediction pipeline” (M6-M60), Task 3.3 “Development of an integrated risk-effect estimation pipeline” (M6-M60), and Task 3.4 “Development of a pipeline for disease trajectory analysis” (M12-M36).

BackgroundRecent evidence suggests that there is often substantial variation in the benefits and harms across a trial population. We aimed to identify regression modeling approaches that assess heterogeneity of treatment effect within a randomized clinical trial.MethodsWe performed a literature review using a broad search strategy, complemented by suggestions of a technical expert panel.ResultsThe approaches are classified into 3 categories: 1) Risk-based methods (11 papers) use only prognostic factors to define patient subgroups, relying on the mathematical dependency of the absolute risk difference on baseline risk; 2) Treatment effect modeling methods (9 papers) use both prognostic factors and treatment effect modifiers to explore characteristics that interact with the effects of therapy on a relative scale. These methods couple data-driven subgroup identification with approaches to prevent overfitting, such as penalization or use of separate data sets for subgroup identification and effect estimation. 3) Optimal treatment regime methods (12 papers) focus primarily on treatment effect modifiers to classify the trial population into those who benefit from treatment and those who do not. Finally, we also identified papers which describe model evaluation methods (4 papers).ConclusionThree classes of approaches were identified to assess heterogeneity of treatment effect. Methodological research, including both simulations and empirical evaluations, is required to compare the available methods in different settings and to derive well-informed guidance for their application in RCT analysis.Key messagesHeterogeneity of treatment effect refers to the non-random variation in the direction or magnitude of a treatment effect for individuals within a population.A large number of regression-based predictive approaches to the analysis of treatment effect heterogeneity exists, which can be divided into three broad classes based on if they incorporate: prognostic factors (risk-based methods); treatment effect modifiers (optimal treatment regime methods); or both (treatment effect modeling methods).Simulations and empirical evaluations are required to compare the available methods in different settings and to derive well-informed guidance for their application in RCT analysis.

There are large opportunities to use the massive amount of observational data in Europe for personalized decision-making. However, the lack of inter-operability of the data sources makes this a challenging task. The differences in structure (syntactic inter-operability) and terminologies systems (semantic inter-operability) make the development of standardized analytical pipelines cumbersome. The European Health Data and Evidence Network (EHDEN) project is addressing this by standardizing a large amount of European data sources to the OMOP Common Data Model (CDM). The goal of WP3 “Personalized Medicine” is to establish a standardized process to enable personalized decision-making that can be utilized for multiple outcomes of interest and can be applied to observational healthcare data from any patient subpopulation. This first report introduces the analytical pipelines for Patient-Level Prediction and Population-Level Effect Estimation that are being developed by the Observational Health Data Sciences and Informatics (OHDSI) community. This work includes large contributions of EHDEN consortium members. Furthermore, we discuss our development of a pipeline for Risk Stratified Effect Estimation to assess heterogeneity of treatment effect. The analytical pipelines have already been applied successfully in EHDEN in study-a-thons that try to answer important clinical questions in a short timeframe with a dedicated team of clinicians and data scientists. This work falls under Task 3.2. “Development of an integrated patient-level prediction pipeline (M6-M60) and Task 3.3 “Development of an integrated risk-effect estimation pipeline (M6-M60).

The Personalized Medicine Work Package (WP3) in the European Health Data and Evidence Network (EHDEN) project, aims to further develop the analytical pipelines for Patient-Level Prediction and Population-Level Effect Estimation from the Observational Health Data Sciences and Informatics (OHDSI) initiative and will develop a pipeline for assessing Disease Trajectories. This first deliverable of WP3 describes a first selection of use cases in these three domains that will drive the development of the EHDEN infrastructure. This includes methodological research, but also developments for the dissemination of the results in the form of dashboard. Moreover, the use case will demonstrate the power of the data standardisation through the OMOP Common Data Model by executing studies in multiple clinical domains. WP3 has already been able to apply the pipelines in study-a-thons, has presented results in high-impact conferences, and has submitted several publications on these topics. This deliverable also briefly describes these results.

Aim: One of the aims of the Observation Health Data Sciences and Informatics (OHDSI) initiative is population-level treatment effect estimation in large observational databases. Since treatment effects are well-known to vary across groups of patients with different baseline risk, we aimed to extend the OHDSI methods library with a framework for risk-based assessment of treatment effect heterogeneity. Materials and Methods: The proposed framework consists of five steps: 1) definition of the problem, i.e. the population, the treatment, the comparator and the outcome(s) of interest; 2) identification of relevant databases; 3) development of a prediction model for the outcome(s) of interest; 4) estimation of propensity scores within strata of predicted risk and estimation of relative and absolute treatment effect within strata of predicted risk; 5) evaluation and presentation of results. Results: We demonstrate our framework by evaluating heterogeneity of the effect of angiotensin-converting enzyme (ACE) inhibitors versus beta blockers on a set of 9 outcomes of interest across three observational databases. With increasing risk of acute myocardial infarction we observed increasing absolute benefits, i.e. from -0.03% to 0.54% in the lowest to highest risk groups. Cough-related absolute harms decreased from 4.1% to 2.6%. Conclusions: The proposed framework may be useful for the evaluation of heterogeneity of treatment effect on observational data that are mapped to the OMOP Common Data Model. The proof of concept study demonstrates its feasibility in large observational data. Further insights may arise by application to safety and effectiveness questions across the global data network.

There are large opportunities to use the massive amount of observational data in Europe for personalized decision-making. However, the lack of inter-operability of the data sources makes this a challenging task. The differences in structure (syntactic inter-operability) and terminology systems (semantic inter-operability) make the development of standardized analytical pipelines cumbersome. The European Health Data and Evidence Network (EHDEN) project is addressing this by standardizing a large amount of European data sources to the OMOP Common Data Model (CDM). The goal of WP3 “Personalized Medicine” is to establish a standardized process to enable personalized decision-making that can be utilized for multiple outcomes of interest and can be applied to observational healthcare data from any patient subpopulation. In the first report on WP3 activities, we introduced the analytical pipelines for Patient-Level Prediction and Population-Level Effect Estimation. Furthermore, we discussed our initial work for the development of a pipeline for Risk Stratified Effect Estimation to assess heterogeneity of treatment effect. The current second report provides an update on the work done in the second year. This includes an overview of use cases in which the analytical pipelines have been applied and describes the advances made in methodological research, the start of a natural language processing pipeline, and work done to develop a pipeline for disease trajectories. This work falls under Task 3.2. “Development of an integrated patient-level prediction pipeline” (M6-M60), Task 3.3 “Development of an integrated risk-effect estimation pipeline” (M6-M60), and Task 3.4 “Development of a pipeline for disease trajectory analysis” (M12-M36).