FAQs

The creation of SCAs typically involves several steps:

1. Identifying and accessing relevant real-world data sources

2. Curating and cleaning the data to ensure quality and completeness

3. Selecting a patient cohort that matches the characteristics of the study population

4. Applying statistical methods such as propensity score matching to ensure comparability

5. Conducting rigorous analyses to validate the synthetic control arm

Common sources of real-world data for SCAs include:

Electronic Health Records (EHRs)

Insurance claims databases

Disease-specific registries

National health databases

Patient-reported outcome measures

Some potential advantages of SCAs include:

Reducing the number of patients needed for the control group

Potentially shortening trial duration

Allowing more patients to receive the experimental treatment

Providing real-world context to clinical trial results

Enabling trials in rare diseases or emergency situations where control groups may be unfeasible

Key challenges include:

Ensuring data quality and completeness in real-world data sources

Addressing potential biases in the data

Achieving appropriate matching between the synthetic and treatment arms

Handling missing data

Meeting regulatory requirements for acceptability of SCAs

Regulatory agencies such as the FDA and EMA are increasingly open to the use of SCAs, particularly in certain situations like rare disease trials or where randomization is challenging. However, they require rigorous methodology, high-quality data, and thorough documentation. Agencies have issued guidance documents outlining their expectations for the use of real-world evidence in clinical trials, including as synthetic control arms.

Refer: https://www.fda.gov/regulatory-information/search-fda-guidance-documents/considerations-design-and-conduct-externally-controlled-trials-drug-and-biological-products

Common statistical methods include:

Propensity score matching

Inverse probability of treatment weighting

Doubly robust estimation

Bayesian hierarchical modeling

Machine learning approaches for patient matching and outcome prediction

SCAs may be particularly valuable in:

Rare disease trials where patient recruitment is challenging

Oncology trials, especially for aggressive cancers

Situations where randomization to a placebo or standard of care may be unethical

Studies of emergency interventions (e.g., pandemic situations)

Long-term safety studies

Validity assessment typically involves:

Comparing baseline characteristics between the SCA and the treatment arm

Conducting sensitivity analyses to test the robustness of results

Assessing the plausibility of outcomes in the SCA compared to historical data

Evaluating the completeness and quality of the underlying real-world data

Potentially validating the SCA approach using data from completed randomized controlled trials

Important limitations to consider include:

Potential for unmeasured confounding factors

Challenges in accounting for differences in care patterns across different healthcare settings

Difficulty in blinding investigators to the nature of the control arm

Potential skepticism from some stakeholders about the reliability of non-randomized comparisons

Limitations in available real-world data for certain outcomes or patient populations