By Exploristics CDSO, Kimberley Hacquoil
The FDA have recently published a draft guidance for the design and conduct of externally controlled trials which specifically focuses on the use of patient level data as a full external control arm. This external data can be from other clinical trials or real-world data (RWD) sources e.g., registries, electronic health records, or medical claims. Using patient level data as external controls, rather than summary level data, opens up the opportunities to harness the power of patient level data. This would be particularly powerful in situations where standard practice is impractical due to operational constraints, such as the difficulty to recruit participants rare and orphan diseases. However, there are additional considerations which need to be addressed when designing and analysing such trials.
Design Considerations
Since there will often be a lack of blinding and randomisation for external control arms, a key focus for the FDA guidance is to ensure that at the design stage any bias is reduced, and treatment arms are considered similar for robust comparisons. Finalising the study protocol and prespecifying plans regarding how to measure and analyse data, prior to initiating the external controlled trial is essential. This will increase confidence in the interpretability of the results.
It may be that blinding, confounding factors, and other sources of bias cannot be completely eliminated, however, an assessment of the impact and methods explored to reduce this are critical. For example, a lack of blinding can provide challenges when considering certain (more subjective) outcomes and so it’s advised that methods such as re-adjudication of outcomes by a blinded independent central reviewer be considered. Simulation at the design stage will also be critical to inform decision-making and justification of the design to utilise external control arms. It will support companies to identify the best study design options and investigate different assumptions and scenarios when using external control arms.
Data Considerations
Utilising data from another clinical trial rather than data from routine clinical care can have advantages since it will be part of a more rigorous protocol-based collection. However, regardless of the source of data, it’s important to address the following to understand and manage potential treats to the validity of externally controlled trials:
- Clinical care changes over time
- Variations based on geographic region and healthcare systems
- Criteria used to establish diagnosis
- Similarity of prognostic indicators
- Attributes of the treatment of interest
- Other treatment related factors (e.g., previous treatments, concomitant medications, predictive biomarkers)
- Follow-up periods and designation of index date
- Relevance of intercurrent events
- Reliable and consistent measurement of outcomes
- Extent of and reasons for missing data
Analysis Considerations
The FDA does not recommend a particular approach to analysing data from externally controlled trials as no single statistical or analytical method is suitable for all situations. As with all clinical trials, the FDA require a statistical analysis plan which pre-specifies analyses with justification of analytical methods, statistical power, sample size and control of type I error. It’s important that decisions regarding the design and analysis for externally controlled trials are done so in a blinded way to the observed data (especially key for historical clinical trials which may have already reported results), although evaluation of key variables and missing data are permitted to ensure sufficient and relevant data.
Analytical methods should include a strategy and assumptions for dealing with missing data and misclassification which can be more common when working with RWD sources. Additional sensitivity analyses can be used to test the vulnerability of trial results and bias to the assumptions in the analysis plan.
Regulatory Considerations
As always, it’s essential to engage early with the regulators and justify aspects related to the proposed study design and statistical analyses. Within the guidance it also highlights the additional elements regarding the proposed data sources and why they are a good fit for use as an external control. It’s also key for sponsors to address the FDAs expectations for the submission of data, especially if the sponsor doesn’t “own” the external data. For example, if the FDA require access to source documents and source data for external control arm of an FDA inspection, how will the sponsor ensure this is possible?
Worth it?
For decades the FDA has recognised the potential value for other types of controls for submissions including historical controls. The new guidance highlights that the FDA are taking steps to broaden the thinking around the use of external patient level data in clinical trials. With so much patient data out there, and with the challenges within drug development, we need to think of robust, unbiased, and efficient ways to better utilise this vast information. There are also further opportunities to supplement control arms with external data (which is not covered in this guidance) which may overcome some of the constraints and concerns regarding bias. Indeed, sponsors may be more willing to adopt a hybrid approach as an initial step towards implementing external control arms.
This guidance helps companies who are seeking alternative approaches to drug development with the ability to streamline plans and utilise other sources of information in a formal way for regulatory submissions. It’s relevant to companies of all sizes, from smaller biotechs to large pharmaceuticals, and may provide cost effective ways to speed up development without risking regulatory success.
I would love to see future regulatory advice on this topic and more regulatory submissions which use patient level data as an external control arm.
Read more:
Synthetic control arms, making the leap beyond the gold standard
Are synthetic control arms the future standard in clinical trials?
It’s not rare to have a rare disease
Why thinking big for small populations is transforming rare disease drug development
Going synthetic with real-world data
Watch more: