Case Study
Construction of a synthetic control arm using real-world data electronic health records
The Data Science team provide advanced analytics using proprietary algorithms, tools, machine learning and predictive modelling approaches to extract insights and synthesize evidence that can be used to inform the assumptions, strategy, and direction of any clinical trial program.
The Challenge
A biotechnology company was investigating a newly developed technology which aimed to enable radiotherapy to kill more tumour cells while minimizing the effects on healthy tissue. They had access to an extensive database of real-world healthcare data and wanted to generate evidence in support of the development of a synthetic control arm for their clinical program in a specific oncology subpopulation, with specific prerequisites for prior treatment regime and disease severity. The company wished to consider if they could use real-world data (RWD) in a synthetic control arm to replace enrolling patients.
Why Synthetic Control Arms?
There is a growing demand for the use of real-world data to support regulatory approval in contexts where a randomised control trial would not be feasible, especially when the use of a placebo would inhibit participant recruitment and retention. By eliminating the need to enrol control participants, synthetic control arms can increase efficiency, reduce trial costs, and accelerate much needed drugs to market.
The Approach
The Data Science team conducted a structured review and evaluation of the data which included data synthesis, data wrangling and analysis to identify patients derived from the database who aligned with the inclusion and exclusion criteria and baseline characteristics of patients taking part in a clinical trial.
The key objectives were to:
- Characterise the available real-world data including key outcomes in subgroups.
- Assess the utility and strategy for a synthetic control group.
Characterisation of real-world data
The RWD registry contained thousands of entries for the main oncology disease area, however the synthetic control arm required a cohort with specific eligibility criteria including matching to:
- Clinical trial timelines
- A treatment regimen of named radiotherapy and chemotherapy in a specific order
- Specific location of disease
- Disease severity
Therefore, the search was filtered using these criteria and a RWD cohort was constructed using the method outlined in Figure 1:
Figure 1. Methodology for generating a RWD cohort.
Feasibility of constructing a synthetic control arm
Once a RWD cohort had been constructed, the following process (Figure 2) was employed in assessing the feasibility of a synthetic control arm for clinical development and clinical trial optimisation:
Figure 2. Using RWD to generate real-world evidence (RWE) in clinical trials.
FDA guidance on the use of RWE to support regulatory decision-making for medical devices (1) emphasises the importance of having a pre-defined common set of data elements, a common definitional framework (i.e., data dictionary), and pre-specified time intervals for data element collection and outcome analyses.
The Data Science team:
- derived and documented elements including key aspects related to data missingness, data integrity and alignment, and derivation of clinical time-to-event outcomes such as progression free survival.
- employed a rigorous, pre-specified procedure of patient selection from the RWD without knowledge of the single arm trial results to reduce any potential form of bias.
The Approach
The Data Science team conducted a structured review and evaluation of the data which included data synthesis, data wrangling and analysis to identify patients derived from the database who aligned with the inclusion and exclusion criteria and baseline characteristics of patients taking part in a clinical trial.
The key objectives were to:
- Characterise the available real-world data including key outcomes in subgroups.
- Assess the utility and strategy for a synthetic control group.
Characterisation of real-world data
The RWD registry contained thousands of entries for the main oncology disease area, however the synthetic control arm required a cohort with specific eligibility criteria including matching to:
- Clinical trial timelines
- A treatment regimen of named radiotherapy and chemotherapy in a specific order
- Specific location of disease
- Disease severity
Therefore, the search was filtered using these criteria and a RWD cohort was constructed using the method outlined in Figure 1:
Figure 1. Methodology for generating a RWD cohort.
Feasibility of constructing a synthetic control arm
Once a RWD cohort had been constructed, the following process (Figure 2) was employed in assessing the feasibility of a synthetic control arm for clinical development and clinical trial optimisation:
Figure 2. Using RWD to generate real-world evidence (RWE) in clinical trials.
FDA guidance on the use of RWE to support regulatory decision-making for medical devices (1) emphasises the importance of having a pre-defined common set of data elements, a common definitional framework (i.e., data dictionary), and pre-specified time intervals for data element collection and outcome analyses.
The Data Science team:
- derived and documented elements including key aspects related to data missingness, data integrity and alignment, and derivation of clinical time-to-event outcomes such as progression free survival.
- employed a rigorous, pre-specified procedure of patient selection from the RWD without knowledge of the single arm trial results to reduce any potential form of bias.
The Results
Collation of a real-world data cohort
As each of the prespecified search criteria required for the study was applied to the RWD registry population the pool of eligible candidates narrowed from thousands to low hundreds and then to tens. These were used to construct a RWD cohort.
Feasibility of using a synthetic control arm
Finding external control data that matches the specifications of the single arm trial can be challenging, especially in small populations and rare diseases when specific eligibility criteria are required, for example, prior treatment regimen or disease severity.
Sufficient patient-level data was identified and filtered from the RWD registry for use in a synthetic control arm. The synthetic control arm was successfully validated for a clinical trial investigating a particular oncology subpopulation who had a specific disease and treatment history.
The Impact
The company investigating a candidate therapeutic for a disease for which there are usually small populations were able to:
- Use RWD in a synthetic control arm for their study to replace enrolling patients
- Generate sufficient data for their study despite the small target population
- Save recruitment time, costs and patient burden
References
Let’s talk!
If you’d like to discuss this case study further or learn more on how our technology enabled services can support your development project, please contact our VP of Sales & Marketing, Abbas Shivji, at abbas.shivji@exploristics.com or book a call.